2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
111 * transaction start -> page lock(s) -> i_data_sem (rw)
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #define IS_EXT2_SB(sb) (0)
130 static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141 static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
144 if (!ext4_has_feature_metadata_csum(sb))
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
159 return cpu_to_le32(csum);
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
165 if (!ext4_has_metadata_csum(sb))
168 return es->s_checksum == ext4_superblock_csum(sb, es);
171 void ext4_superblock_csum_set(struct super_block *sb)
173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
175 if (!ext4_has_metadata_csum(sb))
178 es->s_checksum = ext4_superblock_csum(sb, es);
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
185 ret = kmalloc(size, flags | __GFP_NOWARN);
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
195 ret = kzalloc(size, flags | __GFP_NOWARN);
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
220 return le32_to_cpu(bg->bg_inode_table_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257 void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265 void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273 void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281 void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289 void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297 void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305 void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
314 static void __save_error_info(struct super_block *sb, const char *func,
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320 if (bdev_read_only(sb->s_bdev))
322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
335 * Start the daily error reporting function if it hasn't been
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340 le32_add_cpu(&es->s_error_count, 1);
343 static void save_error_info(struct super_block *sb, const char *func,
346 __save_error_info(sb, func, line);
347 if (!bdev_read_only(sb->s_bdev))
348 ext4_commit_super(sb, 1);
352 * The del_gendisk() function uninitializes the disk-specific data
353 * structures, including the bdi structure, without telling anyone
354 * else. Once this happens, any attempt to call mark_buffer_dirty()
355 * (for example, by ext4_commit_super), will cause a kernel OOPS.
356 * This is a kludge to prevent these oops until we can put in a proper
357 * hook in del_gendisk() to inform the VFS and file system layers.
359 static int block_device_ejected(struct super_block *sb)
361 struct inode *bd_inode = sb->s_bdev->bd_inode;
362 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
364 return bdi->dev == NULL;
367 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
369 struct super_block *sb = journal->j_private;
370 struct ext4_sb_info *sbi = EXT4_SB(sb);
371 int error = is_journal_aborted(journal);
372 struct ext4_journal_cb_entry *jce;
374 BUG_ON(txn->t_state == T_FINISHED);
375 spin_lock(&sbi->s_md_lock);
376 while (!list_empty(&txn->t_private_list)) {
377 jce = list_entry(txn->t_private_list.next,
378 struct ext4_journal_cb_entry, jce_list);
379 list_del_init(&jce->jce_list);
380 spin_unlock(&sbi->s_md_lock);
381 jce->jce_func(sb, jce, error);
382 spin_lock(&sbi->s_md_lock);
384 spin_unlock(&sbi->s_md_lock);
387 /* Deal with the reporting of failure conditions on a filesystem such as
388 * inconsistencies detected or read IO failures.
390 * On ext2, we can store the error state of the filesystem in the
391 * superblock. That is not possible on ext4, because we may have other
392 * write ordering constraints on the superblock which prevent us from
393 * writing it out straight away; and given that the journal is about to
394 * be aborted, we can't rely on the current, or future, transactions to
395 * write out the superblock safely.
397 * We'll just use the jbd2_journal_abort() error code to record an error in
398 * the journal instead. On recovery, the journal will complain about
399 * that error until we've noted it down and cleared it.
402 static void ext4_handle_error(struct super_block *sb)
404 if (sb->s_flags & MS_RDONLY)
407 if (!test_opt(sb, ERRORS_CONT)) {
408 journal_t *journal = EXT4_SB(sb)->s_journal;
410 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
412 jbd2_journal_abort(journal, -EIO);
414 if (test_opt(sb, ERRORS_RO)) {
415 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
417 * Make sure updated value of ->s_mount_flags will be visible
418 * before ->s_flags update
421 sb->s_flags |= MS_RDONLY;
423 if (test_opt(sb, ERRORS_PANIC)) {
424 if (EXT4_SB(sb)->s_journal &&
425 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
427 panic("EXT4-fs (device %s): panic forced after error\n",
432 #define ext4_error_ratelimit(sb) \
433 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
436 void __ext4_error(struct super_block *sb, const char *function,
437 unsigned int line, const char *fmt, ...)
439 struct va_format vaf;
442 if (ext4_error_ratelimit(sb)) {
447 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
448 sb->s_id, function, line, current->comm, &vaf);
451 save_error_info(sb, function, line);
452 ext4_handle_error(sb);
455 void __ext4_error_inode(struct inode *inode, const char *function,
456 unsigned int line, ext4_fsblk_t block,
457 const char *fmt, ...)
460 struct va_format vaf;
461 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
463 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
464 es->s_last_error_block = cpu_to_le64(block);
465 if (ext4_error_ratelimit(inode->i_sb)) {
470 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
471 "inode #%lu: block %llu: comm %s: %pV\n",
472 inode->i_sb->s_id, function, line, inode->i_ino,
473 block, current->comm, &vaf);
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476 "inode #%lu: comm %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 current->comm, &vaf);
481 save_error_info(inode->i_sb, function, line);
482 ext4_handle_error(inode->i_sb);
485 void __ext4_error_file(struct file *file, const char *function,
486 unsigned int line, ext4_fsblk_t block,
487 const char *fmt, ...)
490 struct va_format vaf;
491 struct ext4_super_block *es;
492 struct inode *inode = file_inode(file);
493 char pathname[80], *path;
495 es = EXT4_SB(inode->i_sb)->s_es;
496 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
497 if (ext4_error_ratelimit(inode->i_sb)) {
498 path = file_path(file, pathname, sizeof(pathname));
506 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
507 "block %llu: comm %s: path %s: %pV\n",
508 inode->i_sb->s_id, function, line, inode->i_ino,
509 block, current->comm, path, &vaf);
512 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
513 "comm %s: path %s: %pV\n",
514 inode->i_sb->s_id, function, line, inode->i_ino,
515 current->comm, path, &vaf);
518 save_error_info(inode->i_sb, function, line);
519 ext4_handle_error(inode->i_sb);
522 const char *ext4_decode_error(struct super_block *sb, int errno,
529 errstr = "Corrupt filesystem";
532 errstr = "Filesystem failed CRC";
535 errstr = "IO failure";
538 errstr = "Out of memory";
541 if (!sb || (EXT4_SB(sb)->s_journal &&
542 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
543 errstr = "Journal has aborted";
545 errstr = "Readonly filesystem";
548 /* If the caller passed in an extra buffer for unknown
549 * errors, textualise them now. Else we just return
552 /* Check for truncated error codes... */
553 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
562 /* __ext4_std_error decodes expected errors from journaling functions
563 * automatically and invokes the appropriate error response. */
565 void __ext4_std_error(struct super_block *sb, const char *function,
566 unsigned int line, int errno)
571 /* Special case: if the error is EROFS, and we're not already
572 * inside a transaction, then there's really no point in logging
574 if (errno == -EROFS && journal_current_handle() == NULL &&
575 (sb->s_flags & MS_RDONLY))
578 if (ext4_error_ratelimit(sb)) {
579 errstr = ext4_decode_error(sb, errno, nbuf);
580 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
581 sb->s_id, function, line, errstr);
584 save_error_info(sb, function, line);
585 ext4_handle_error(sb);
589 * ext4_abort is a much stronger failure handler than ext4_error. The
590 * abort function may be used to deal with unrecoverable failures such
591 * as journal IO errors or ENOMEM at a critical moment in log management.
593 * We unconditionally force the filesystem into an ABORT|READONLY state,
594 * unless the error response on the fs has been set to panic in which
595 * case we take the easy way out and panic immediately.
598 void __ext4_abort(struct super_block *sb, const char *function,
599 unsigned int line, const char *fmt, ...)
601 struct va_format vaf;
604 save_error_info(sb, function, line);
608 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
609 sb->s_id, function, line, &vaf);
612 if ((sb->s_flags & MS_RDONLY) == 0) {
613 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
614 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
616 * Make sure updated value of ->s_mount_flags will be visible
617 * before ->s_flags update
620 sb->s_flags |= MS_RDONLY;
621 if (EXT4_SB(sb)->s_journal)
622 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
623 save_error_info(sb, function, line);
625 if (test_opt(sb, ERRORS_PANIC)) {
626 if (EXT4_SB(sb)->s_journal &&
627 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
629 panic("EXT4-fs panic from previous error\n");
633 void __ext4_msg(struct super_block *sb,
634 const char *prefix, const char *fmt, ...)
636 struct va_format vaf;
639 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
645 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
649 #define ext4_warning_ratelimit(sb) \
650 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
653 void __ext4_warning(struct super_block *sb, const char *function,
654 unsigned int line, const char *fmt, ...)
656 struct va_format vaf;
659 if (!ext4_warning_ratelimit(sb))
665 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
666 sb->s_id, function, line, &vaf);
670 void __ext4_warning_inode(const struct inode *inode, const char *function,
671 unsigned int line, const char *fmt, ...)
673 struct va_format vaf;
676 if (!ext4_warning_ratelimit(inode->i_sb))
682 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
683 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
684 function, line, inode->i_ino, current->comm, &vaf);
688 void __ext4_grp_locked_error(const char *function, unsigned int line,
689 struct super_block *sb, ext4_group_t grp,
690 unsigned long ino, ext4_fsblk_t block,
691 const char *fmt, ...)
695 struct va_format vaf;
697 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
699 es->s_last_error_ino = cpu_to_le32(ino);
700 es->s_last_error_block = cpu_to_le64(block);
701 __save_error_info(sb, function, line);
703 if (ext4_error_ratelimit(sb)) {
707 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
708 sb->s_id, function, line, grp);
710 printk(KERN_CONT "inode %lu: ", ino);
712 printk(KERN_CONT "block %llu:",
713 (unsigned long long) block);
714 printk(KERN_CONT "%pV\n", &vaf);
718 if (test_opt(sb, ERRORS_CONT)) {
719 ext4_commit_super(sb, 0);
723 ext4_unlock_group(sb, grp);
724 ext4_commit_super(sb, 1);
725 ext4_handle_error(sb);
727 * We only get here in the ERRORS_RO case; relocking the group
728 * may be dangerous, but nothing bad will happen since the
729 * filesystem will have already been marked read/only and the
730 * journal has been aborted. We return 1 as a hint to callers
731 * who might what to use the return value from
732 * ext4_grp_locked_error() to distinguish between the
733 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
734 * aggressively from the ext4 function in question, with a
735 * more appropriate error code.
737 ext4_lock_group(sb, grp);
741 void ext4_update_dynamic_rev(struct super_block *sb)
743 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
745 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
749 "updating to rev %d because of new feature flag, "
750 "running e2fsck is recommended",
753 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
754 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
755 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
756 /* leave es->s_feature_*compat flags alone */
757 /* es->s_uuid will be set by e2fsck if empty */
760 * The rest of the superblock fields should be zero, and if not it
761 * means they are likely already in use, so leave them alone. We
762 * can leave it up to e2fsck to clean up any inconsistencies there.
767 * Open the external journal device
769 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
771 struct block_device *bdev;
772 char b[BDEVNAME_SIZE];
774 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
780 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
781 __bdevname(dev, b), PTR_ERR(bdev));
786 * Release the journal device
788 static void ext4_blkdev_put(struct block_device *bdev)
790 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
793 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
795 struct block_device *bdev;
796 bdev = sbi->journal_bdev;
798 ext4_blkdev_put(bdev);
799 sbi->journal_bdev = NULL;
803 static inline struct inode *orphan_list_entry(struct list_head *l)
805 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
808 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
812 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
813 le32_to_cpu(sbi->s_es->s_last_orphan));
815 printk(KERN_ERR "sb_info orphan list:\n");
816 list_for_each(l, &sbi->s_orphan) {
817 struct inode *inode = orphan_list_entry(l);
819 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
820 inode->i_sb->s_id, inode->i_ino, inode,
821 inode->i_mode, inode->i_nlink,
826 static void ext4_put_super(struct super_block *sb)
828 struct ext4_sb_info *sbi = EXT4_SB(sb);
829 struct ext4_super_block *es = sbi->s_es;
830 struct buffer_head **group_desc;
831 struct flex_groups **flex_groups;
835 ext4_unregister_li_request(sb);
836 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
838 flush_workqueue(sbi->rsv_conversion_wq);
839 destroy_workqueue(sbi->rsv_conversion_wq);
841 if (sbi->s_journal) {
842 aborted = is_journal_aborted(sbi->s_journal);
843 err = jbd2_journal_destroy(sbi->s_journal);
844 sbi->s_journal = NULL;
845 if ((err < 0) && !aborted)
846 ext4_abort(sb, "Couldn't clean up the journal");
849 ext4_unregister_sysfs(sb);
850 ext4_es_unregister_shrinker(sbi);
851 del_timer_sync(&sbi->s_err_report);
852 ext4_release_system_zone(sb);
854 ext4_ext_release(sb);
856 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
857 ext4_clear_feature_journal_needs_recovery(sb);
858 es->s_state = cpu_to_le16(sbi->s_mount_state);
860 if (!(sb->s_flags & MS_RDONLY))
861 ext4_commit_super(sb, 1);
864 group_desc = rcu_dereference(sbi->s_group_desc);
865 for (i = 0; i < sbi->s_gdb_count; i++)
866 brelse(group_desc[i]);
868 flex_groups = rcu_dereference(sbi->s_flex_groups);
870 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
871 kvfree(flex_groups[i]);
875 percpu_counter_destroy(&sbi->s_freeclusters_counter);
876 percpu_counter_destroy(&sbi->s_freeinodes_counter);
877 percpu_counter_destroy(&sbi->s_dirs_counter);
878 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
879 percpu_free_rwsem(&sbi->s_writepages_rwsem);
882 for (i = 0; i < EXT4_MAXQUOTAS; i++)
883 kfree(sbi->s_qf_names[i]);
886 /* Debugging code just in case the in-memory inode orphan list
887 * isn't empty. The on-disk one can be non-empty if we've
888 * detected an error and taken the fs readonly, but the
889 * in-memory list had better be clean by this point. */
890 if (!list_empty(&sbi->s_orphan))
891 dump_orphan_list(sb, sbi);
892 J_ASSERT(list_empty(&sbi->s_orphan));
894 sync_blockdev(sb->s_bdev);
895 invalidate_bdev(sb->s_bdev);
896 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
898 * Invalidate the journal device's buffers. We don't want them
899 * floating about in memory - the physical journal device may
900 * hotswapped, and it breaks the `ro-after' testing code.
902 sync_blockdev(sbi->journal_bdev);
903 invalidate_bdev(sbi->journal_bdev);
904 ext4_blkdev_remove(sbi);
906 if (sbi->s_mb_cache) {
907 ext4_xattr_destroy_cache(sbi->s_mb_cache);
908 sbi->s_mb_cache = NULL;
911 kthread_stop(sbi->s_mmp_tsk);
912 sb->s_fs_info = NULL;
914 * Now that we are completely done shutting down the
915 * superblock, we need to actually destroy the kobject.
917 kobject_put(&sbi->s_kobj);
918 wait_for_completion(&sbi->s_kobj_unregister);
919 if (sbi->s_chksum_driver)
920 crypto_free_shash(sbi->s_chksum_driver);
921 kfree(sbi->s_blockgroup_lock);
925 static struct kmem_cache *ext4_inode_cachep;
928 * Called inside transaction, so use GFP_NOFS
930 static struct inode *ext4_alloc_inode(struct super_block *sb)
932 struct ext4_inode_info *ei;
934 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
938 ei->vfs_inode.i_version = 1;
939 spin_lock_init(&ei->i_raw_lock);
940 INIT_LIST_HEAD(&ei->i_prealloc_list);
941 spin_lock_init(&ei->i_prealloc_lock);
942 ext4_es_init_tree(&ei->i_es_tree);
943 rwlock_init(&ei->i_es_lock);
944 INIT_LIST_HEAD(&ei->i_es_list);
947 ei->i_es_shrink_lblk = 0;
948 ei->i_reserved_data_blocks = 0;
949 ei->i_reserved_meta_blocks = 0;
950 ei->i_allocated_meta_blocks = 0;
951 ei->i_da_metadata_calc_len = 0;
952 ei->i_da_metadata_calc_last_lblock = 0;
953 spin_lock_init(&(ei->i_block_reservation_lock));
955 ei->i_reserved_quota = 0;
956 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
959 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
960 spin_lock_init(&ei->i_completed_io_lock);
962 ei->i_datasync_tid = 0;
963 atomic_set(&ei->i_unwritten, 0);
964 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
965 return &ei->vfs_inode;
968 static int ext4_drop_inode(struct inode *inode)
970 int drop = generic_drop_inode(inode);
972 trace_ext4_drop_inode(inode, drop);
976 static void ext4_i_callback(struct rcu_head *head)
978 struct inode *inode = container_of(head, struct inode, i_rcu);
979 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
982 static void ext4_destroy_inode(struct inode *inode)
984 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
985 ext4_msg(inode->i_sb, KERN_ERR,
986 "Inode %lu (%p): orphan list check failed!",
987 inode->i_ino, EXT4_I(inode));
988 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
989 EXT4_I(inode), sizeof(struct ext4_inode_info),
993 call_rcu(&inode->i_rcu, ext4_i_callback);
996 static void init_once(void *foo)
998 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1000 INIT_LIST_HEAD(&ei->i_orphan);
1001 init_rwsem(&ei->xattr_sem);
1002 init_rwsem(&ei->i_data_sem);
1003 init_rwsem(&ei->i_mmap_sem);
1004 inode_init_once(&ei->vfs_inode);
1007 static int __init init_inodecache(void)
1009 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1010 sizeof(struct ext4_inode_info),
1011 0, (SLAB_RECLAIM_ACCOUNT|
1012 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1014 if (ext4_inode_cachep == NULL)
1019 static void destroy_inodecache(void)
1022 * Make sure all delayed rcu free inodes are flushed before we
1026 kmem_cache_destroy(ext4_inode_cachep);
1029 void ext4_clear_inode(struct inode *inode)
1031 invalidate_inode_buffers(inode);
1034 ext4_discard_preallocations(inode);
1035 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1036 if (EXT4_I(inode)->jinode) {
1037 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1038 EXT4_I(inode)->jinode);
1039 jbd2_free_inode(EXT4_I(inode)->jinode);
1040 EXT4_I(inode)->jinode = NULL;
1042 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1043 fscrypt_put_encryption_info(inode, NULL);
1047 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1048 u64 ino, u32 generation)
1050 struct inode *inode;
1053 * Currently we don't know the generation for parent directory, so
1054 * a generation of 0 means "accept any"
1056 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1058 return ERR_CAST(inode);
1059 if (generation && inode->i_generation != generation) {
1061 return ERR_PTR(-ESTALE);
1067 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1068 int fh_len, int fh_type)
1070 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1071 ext4_nfs_get_inode);
1074 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1075 int fh_len, int fh_type)
1077 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1078 ext4_nfs_get_inode);
1081 static int ext4_nfs_commit_metadata(struct inode *inode)
1083 struct writeback_control wbc = {
1084 .sync_mode = WB_SYNC_ALL
1087 trace_ext4_nfs_commit_metadata(inode);
1088 return ext4_write_inode(inode, &wbc);
1092 * Try to release metadata pages (indirect blocks, directories) which are
1093 * mapped via the block device. Since these pages could have journal heads
1094 * which would prevent try_to_free_buffers() from freeing them, we must use
1095 * jbd2 layer's try_to_free_buffers() function to release them.
1097 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1100 journal_t *journal = EXT4_SB(sb)->s_journal;
1102 WARN_ON(PageChecked(page));
1103 if (!page_has_buffers(page))
1106 return jbd2_journal_try_to_free_buffers(journal, page,
1107 wait & ~__GFP_DIRECT_RECLAIM);
1108 return try_to_free_buffers(page);
1111 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1112 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1114 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1115 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1118 static int ext4_key_prefix(struct inode *inode, u8 **key)
1120 *key = EXT4_SB(inode->i_sb)->key_prefix;
1121 return EXT4_SB(inode->i_sb)->key_prefix_size;
1124 static int ext4_prepare_context(struct inode *inode)
1126 return ext4_convert_inline_data(inode);
1129 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1135 /* fs_data is null when internally used. */
1137 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1138 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1141 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1142 ext4_clear_inode_state(inode,
1143 EXT4_STATE_MAY_INLINE_DATA);
1148 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1149 ext4_jbd2_credits_xattr(inode));
1151 return PTR_ERR(handle);
1153 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1154 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1157 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1158 res = ext4_mark_inode_dirty(handle, inode);
1160 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1162 res2 = ext4_journal_stop(handle);
1168 static int ext4_dummy_context(struct inode *inode)
1170 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1173 static unsigned ext4_max_namelen(struct inode *inode)
1175 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1179 static struct fscrypt_operations ext4_cryptops = {
1180 .get_context = ext4_get_context,
1181 .key_prefix = ext4_key_prefix,
1182 .prepare_context = ext4_prepare_context,
1183 .set_context = ext4_set_context,
1184 .dummy_context = ext4_dummy_context,
1185 .is_encrypted = ext4_encrypted_inode,
1186 .empty_dir = ext4_empty_dir,
1187 .max_namelen = ext4_max_namelen,
1190 static struct fscrypt_operations ext4_cryptops = {
1191 .is_encrypted = ext4_encrypted_inode,
1196 static char *quotatypes[] = INITQFNAMES;
1197 #define QTYPE2NAME(t) (quotatypes[t])
1199 static int ext4_write_dquot(struct dquot *dquot);
1200 static int ext4_acquire_dquot(struct dquot *dquot);
1201 static int ext4_release_dquot(struct dquot *dquot);
1202 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1203 static int ext4_write_info(struct super_block *sb, int type);
1204 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1206 static int ext4_quota_off(struct super_block *sb, int type);
1207 static int ext4_quota_on_mount(struct super_block *sb, int type);
1208 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1209 size_t len, loff_t off);
1210 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1211 const char *data, size_t len, loff_t off);
1212 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1213 unsigned int flags);
1214 static int ext4_enable_quotas(struct super_block *sb);
1215 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1217 static struct dquot **ext4_get_dquots(struct inode *inode)
1219 return EXT4_I(inode)->i_dquot;
1222 static const struct dquot_operations ext4_quota_operations = {
1223 .get_reserved_space = ext4_get_reserved_space,
1224 .write_dquot = ext4_write_dquot,
1225 .acquire_dquot = ext4_acquire_dquot,
1226 .release_dquot = ext4_release_dquot,
1227 .mark_dirty = ext4_mark_dquot_dirty,
1228 .write_info = ext4_write_info,
1229 .alloc_dquot = dquot_alloc,
1230 .destroy_dquot = dquot_destroy,
1231 .get_projid = ext4_get_projid,
1232 .get_next_id = ext4_get_next_id,
1235 static const struct quotactl_ops ext4_qctl_operations = {
1236 .quota_on = ext4_quota_on,
1237 .quota_off = ext4_quota_off,
1238 .quota_sync = dquot_quota_sync,
1239 .get_state = dquot_get_state,
1240 .set_info = dquot_set_dqinfo,
1241 .get_dqblk = dquot_get_dqblk,
1242 .set_dqblk = dquot_set_dqblk,
1243 .get_nextdqblk = dquot_get_next_dqblk,
1247 static const struct super_operations ext4_sops = {
1248 .alloc_inode = ext4_alloc_inode,
1249 .destroy_inode = ext4_destroy_inode,
1250 .write_inode = ext4_write_inode,
1251 .dirty_inode = ext4_dirty_inode,
1252 .drop_inode = ext4_drop_inode,
1253 .evict_inode = ext4_evict_inode,
1254 .put_super = ext4_put_super,
1255 .sync_fs = ext4_sync_fs,
1256 .freeze_fs = ext4_freeze,
1257 .unfreeze_fs = ext4_unfreeze,
1258 .statfs = ext4_statfs,
1259 .remount_fs = ext4_remount,
1260 .show_options = ext4_show_options,
1262 .quota_read = ext4_quota_read,
1263 .quota_write = ext4_quota_write,
1264 .get_dquots = ext4_get_dquots,
1266 .bdev_try_to_free_page = bdev_try_to_free_page,
1269 static const struct export_operations ext4_export_ops = {
1270 .fh_to_dentry = ext4_fh_to_dentry,
1271 .fh_to_parent = ext4_fh_to_parent,
1272 .get_parent = ext4_get_parent,
1273 .commit_metadata = ext4_nfs_commit_metadata,
1277 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1278 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1279 Opt_nouid32, Opt_debug, Opt_removed,
1280 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1281 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1282 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1283 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1284 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1285 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1286 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1287 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1288 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1289 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1290 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1291 Opt_lazytime, Opt_nolazytime,
1292 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1293 Opt_inode_readahead_blks, Opt_journal_ioprio,
1294 Opt_dioread_nolock, Opt_dioread_lock,
1295 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1296 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1299 static const match_table_t tokens = {
1300 {Opt_bsd_df, "bsddf"},
1301 {Opt_minix_df, "minixdf"},
1302 {Opt_grpid, "grpid"},
1303 {Opt_grpid, "bsdgroups"},
1304 {Opt_nogrpid, "nogrpid"},
1305 {Opt_nogrpid, "sysvgroups"},
1306 {Opt_resgid, "resgid=%u"},
1307 {Opt_resuid, "resuid=%u"},
1309 {Opt_err_cont, "errors=continue"},
1310 {Opt_err_panic, "errors=panic"},
1311 {Opt_err_ro, "errors=remount-ro"},
1312 {Opt_nouid32, "nouid32"},
1313 {Opt_debug, "debug"},
1314 {Opt_removed, "oldalloc"},
1315 {Opt_removed, "orlov"},
1316 {Opt_user_xattr, "user_xattr"},
1317 {Opt_nouser_xattr, "nouser_xattr"},
1319 {Opt_noacl, "noacl"},
1320 {Opt_noload, "norecovery"},
1321 {Opt_noload, "noload"},
1322 {Opt_removed, "nobh"},
1323 {Opt_removed, "bh"},
1324 {Opt_commit, "commit=%u"},
1325 {Opt_min_batch_time, "min_batch_time=%u"},
1326 {Opt_max_batch_time, "max_batch_time=%u"},
1327 {Opt_journal_dev, "journal_dev=%u"},
1328 {Opt_journal_path, "journal_path=%s"},
1329 {Opt_journal_checksum, "journal_checksum"},
1330 {Opt_nojournal_checksum, "nojournal_checksum"},
1331 {Opt_journal_async_commit, "journal_async_commit"},
1332 {Opt_abort, "abort"},
1333 {Opt_data_journal, "data=journal"},
1334 {Opt_data_ordered, "data=ordered"},
1335 {Opt_data_writeback, "data=writeback"},
1336 {Opt_data_err_abort, "data_err=abort"},
1337 {Opt_data_err_ignore, "data_err=ignore"},
1338 {Opt_offusrjquota, "usrjquota="},
1339 {Opt_usrjquota, "usrjquota=%s"},
1340 {Opt_offgrpjquota, "grpjquota="},
1341 {Opt_grpjquota, "grpjquota=%s"},
1342 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1343 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1344 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1345 {Opt_grpquota, "grpquota"},
1346 {Opt_noquota, "noquota"},
1347 {Opt_quota, "quota"},
1348 {Opt_usrquota, "usrquota"},
1349 {Opt_prjquota, "prjquota"},
1350 {Opt_barrier, "barrier=%u"},
1351 {Opt_barrier, "barrier"},
1352 {Opt_nobarrier, "nobarrier"},
1353 {Opt_i_version, "i_version"},
1355 {Opt_stripe, "stripe=%u"},
1356 {Opt_delalloc, "delalloc"},
1357 {Opt_lazytime, "lazytime"},
1358 {Opt_nolazytime, "nolazytime"},
1359 {Opt_nodelalloc, "nodelalloc"},
1360 {Opt_removed, "mblk_io_submit"},
1361 {Opt_removed, "nomblk_io_submit"},
1362 {Opt_block_validity, "block_validity"},
1363 {Opt_noblock_validity, "noblock_validity"},
1364 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1365 {Opt_journal_ioprio, "journal_ioprio=%u"},
1366 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1367 {Opt_auto_da_alloc, "auto_da_alloc"},
1368 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1369 {Opt_dioread_nolock, "dioread_nolock"},
1370 {Opt_dioread_lock, "dioread_lock"},
1371 {Opt_discard, "discard"},
1372 {Opt_nodiscard, "nodiscard"},
1373 {Opt_init_itable, "init_itable=%u"},
1374 {Opt_init_itable, "init_itable"},
1375 {Opt_noinit_itable, "noinit_itable"},
1376 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1377 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1378 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1379 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1380 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1381 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1382 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1386 static ext4_fsblk_t get_sb_block(void **data)
1388 ext4_fsblk_t sb_block;
1389 char *options = (char *) *data;
1391 if (!options || strncmp(options, "sb=", 3) != 0)
1392 return 1; /* Default location */
1395 /* TODO: use simple_strtoll with >32bit ext4 */
1396 sb_block = simple_strtoul(options, &options, 0);
1397 if (*options && *options != ',') {
1398 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1402 if (*options == ',')
1404 *data = (void *) options;
1409 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1410 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1411 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1414 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1416 struct ext4_sb_info *sbi = EXT4_SB(sb);
1420 if (sb_any_quota_loaded(sb) &&
1421 !sbi->s_qf_names[qtype]) {
1422 ext4_msg(sb, KERN_ERR,
1423 "Cannot change journaled "
1424 "quota options when quota turned on");
1427 if (ext4_has_feature_quota(sb)) {
1428 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1429 "ignored when QUOTA feature is enabled");
1432 qname = match_strdup(args);
1434 ext4_msg(sb, KERN_ERR,
1435 "Not enough memory for storing quotafile name");
1438 if (sbi->s_qf_names[qtype]) {
1439 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1442 ext4_msg(sb, KERN_ERR,
1443 "%s quota file already specified",
1447 if (strchr(qname, '/')) {
1448 ext4_msg(sb, KERN_ERR,
1449 "quotafile must be on filesystem root");
1452 sbi->s_qf_names[qtype] = qname;
1460 static int clear_qf_name(struct super_block *sb, int qtype)
1463 struct ext4_sb_info *sbi = EXT4_SB(sb);
1465 if (sb_any_quota_loaded(sb) &&
1466 sbi->s_qf_names[qtype]) {
1467 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1468 " when quota turned on");
1471 kfree(sbi->s_qf_names[qtype]);
1472 sbi->s_qf_names[qtype] = NULL;
1477 #define MOPT_SET 0x0001
1478 #define MOPT_CLEAR 0x0002
1479 #define MOPT_NOSUPPORT 0x0004
1480 #define MOPT_EXPLICIT 0x0008
1481 #define MOPT_CLEAR_ERR 0x0010
1482 #define MOPT_GTE0 0x0020
1485 #define MOPT_QFMT 0x0040
1487 #define MOPT_Q MOPT_NOSUPPORT
1488 #define MOPT_QFMT MOPT_NOSUPPORT
1490 #define MOPT_DATAJ 0x0080
1491 #define MOPT_NO_EXT2 0x0100
1492 #define MOPT_NO_EXT3 0x0200
1493 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1494 #define MOPT_STRING 0x0400
1496 static const struct mount_opts {
1500 } ext4_mount_opts[] = {
1501 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1502 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1503 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1504 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1505 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1506 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1507 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1508 MOPT_EXT4_ONLY | MOPT_SET},
1509 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1510 MOPT_EXT4_ONLY | MOPT_CLEAR},
1511 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1512 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1513 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1514 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1515 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1516 MOPT_EXT4_ONLY | MOPT_CLEAR},
1517 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1518 MOPT_EXT4_ONLY | MOPT_CLEAR},
1519 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1520 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1521 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1522 EXT4_MOUNT_JOURNAL_CHECKSUM),
1523 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1524 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1525 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1526 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1527 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1528 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1530 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1532 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1533 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1534 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1535 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1536 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1537 {Opt_commit, 0, MOPT_GTE0},
1538 {Opt_max_batch_time, 0, MOPT_GTE0},
1539 {Opt_min_batch_time, 0, MOPT_GTE0},
1540 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1541 {Opt_init_itable, 0, MOPT_GTE0},
1542 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1543 {Opt_stripe, 0, MOPT_GTE0},
1544 {Opt_resuid, 0, MOPT_GTE0},
1545 {Opt_resgid, 0, MOPT_GTE0},
1546 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1547 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1548 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1549 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1550 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1551 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1552 MOPT_NO_EXT2 | MOPT_DATAJ},
1553 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1554 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1555 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1556 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1557 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1559 {Opt_acl, 0, MOPT_NOSUPPORT},
1560 {Opt_noacl, 0, MOPT_NOSUPPORT},
1562 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1563 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1564 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1565 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1567 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1569 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1571 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1572 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1573 MOPT_CLEAR | MOPT_Q},
1574 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1575 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1576 {Opt_offusrjquota, 0, MOPT_Q},
1577 {Opt_offgrpjquota, 0, MOPT_Q},
1578 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1579 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1580 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1581 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1582 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1586 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1587 substring_t *args, unsigned long *journal_devnum,
1588 unsigned int *journal_ioprio, int is_remount)
1590 struct ext4_sb_info *sbi = EXT4_SB(sb);
1591 const struct mount_opts *m;
1597 if (token == Opt_usrjquota)
1598 return set_qf_name(sb, USRQUOTA, &args[0]);
1599 else if (token == Opt_grpjquota)
1600 return set_qf_name(sb, GRPQUOTA, &args[0]);
1601 else if (token == Opt_offusrjquota)
1602 return clear_qf_name(sb, USRQUOTA);
1603 else if (token == Opt_offgrpjquota)
1604 return clear_qf_name(sb, GRPQUOTA);
1608 case Opt_nouser_xattr:
1609 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1612 return 1; /* handled by get_sb_block() */
1614 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1617 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1620 sb->s_flags |= MS_I_VERSION;
1623 sb->s_flags |= MS_LAZYTIME;
1625 case Opt_nolazytime:
1626 sb->s_flags &= ~MS_LAZYTIME;
1630 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1631 if (token == m->token)
1634 if (m->token == Opt_err) {
1635 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1636 "or missing value", opt);
1640 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1641 ext4_msg(sb, KERN_ERR,
1642 "Mount option \"%s\" incompatible with ext2", opt);
1645 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1646 ext4_msg(sb, KERN_ERR,
1647 "Mount option \"%s\" incompatible with ext3", opt);
1651 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1653 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1655 if (m->flags & MOPT_EXPLICIT) {
1656 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1657 set_opt2(sb, EXPLICIT_DELALLOC);
1658 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1659 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1663 if (m->flags & MOPT_CLEAR_ERR)
1664 clear_opt(sb, ERRORS_MASK);
1665 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1666 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1667 "options when quota turned on");
1671 if (m->flags & MOPT_NOSUPPORT) {
1672 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1673 } else if (token == Opt_commit) {
1675 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1676 sbi->s_commit_interval = HZ * arg;
1677 } else if (token == Opt_max_batch_time) {
1678 sbi->s_max_batch_time = arg;
1679 } else if (token == Opt_min_batch_time) {
1680 sbi->s_min_batch_time = arg;
1681 } else if (token == Opt_inode_readahead_blks) {
1682 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1683 ext4_msg(sb, KERN_ERR,
1684 "EXT4-fs: inode_readahead_blks must be "
1685 "0 or a power of 2 smaller than 2^31");
1688 sbi->s_inode_readahead_blks = arg;
1689 } else if (token == Opt_init_itable) {
1690 set_opt(sb, INIT_INODE_TABLE);
1692 arg = EXT4_DEF_LI_WAIT_MULT;
1693 sbi->s_li_wait_mult = arg;
1694 } else if (token == Opt_max_dir_size_kb) {
1695 sbi->s_max_dir_size_kb = arg;
1696 } else if (token == Opt_stripe) {
1697 sbi->s_stripe = arg;
1698 } else if (token == Opt_resuid) {
1699 uid = make_kuid(current_user_ns(), arg);
1700 if (!uid_valid(uid)) {
1701 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1704 sbi->s_resuid = uid;
1705 } else if (token == Opt_resgid) {
1706 gid = make_kgid(current_user_ns(), arg);
1707 if (!gid_valid(gid)) {
1708 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1711 sbi->s_resgid = gid;
1712 } else if (token == Opt_journal_dev) {
1714 ext4_msg(sb, KERN_ERR,
1715 "Cannot specify journal on remount");
1718 *journal_devnum = arg;
1719 } else if (token == Opt_journal_path) {
1721 struct inode *journal_inode;
1726 ext4_msg(sb, KERN_ERR,
1727 "Cannot specify journal on remount");
1730 journal_path = match_strdup(&args[0]);
1731 if (!journal_path) {
1732 ext4_msg(sb, KERN_ERR, "error: could not dup "
1733 "journal device string");
1737 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1739 ext4_msg(sb, KERN_ERR, "error: could not find "
1740 "journal device path: error %d", error);
1741 kfree(journal_path);
1745 journal_inode = d_inode(path.dentry);
1746 if (!S_ISBLK(journal_inode->i_mode)) {
1747 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1748 "is not a block device", journal_path);
1750 kfree(journal_path);
1754 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1756 kfree(journal_path);
1757 } else if (token == Opt_journal_ioprio) {
1759 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1764 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1765 } else if (token == Opt_test_dummy_encryption) {
1766 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1767 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1768 ext4_msg(sb, KERN_WARNING,
1769 "Test dummy encryption mode enabled");
1771 ext4_msg(sb, KERN_WARNING,
1772 "Test dummy encryption mount option ignored");
1774 } else if (m->flags & MOPT_DATAJ) {
1776 if (!sbi->s_journal)
1777 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1778 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1779 ext4_msg(sb, KERN_ERR,
1780 "Cannot change data mode on remount");
1784 clear_opt(sb, DATA_FLAGS);
1785 sbi->s_mount_opt |= m->mount_opt;
1788 } else if (m->flags & MOPT_QFMT) {
1789 if (sb_any_quota_loaded(sb) &&
1790 sbi->s_jquota_fmt != m->mount_opt) {
1791 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1792 "quota options when quota turned on");
1795 if (ext4_has_feature_quota(sb)) {
1796 ext4_msg(sb, KERN_INFO,
1797 "Quota format mount options ignored "
1798 "when QUOTA feature is enabled");
1801 sbi->s_jquota_fmt = m->mount_opt;
1803 } else if (token == Opt_dax) {
1804 #ifdef CONFIG_FS_DAX
1805 ext4_msg(sb, KERN_WARNING,
1806 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1807 sbi->s_mount_opt |= m->mount_opt;
1809 ext4_msg(sb, KERN_INFO, "dax option not supported");
1812 } else if (token == Opt_data_err_abort) {
1813 sbi->s_mount_opt |= m->mount_opt;
1814 } else if (token == Opt_data_err_ignore) {
1815 sbi->s_mount_opt &= ~m->mount_opt;
1819 if (m->flags & MOPT_CLEAR)
1821 else if (unlikely(!(m->flags & MOPT_SET))) {
1822 ext4_msg(sb, KERN_WARNING,
1823 "buggy handling of option %s", opt);
1828 sbi->s_mount_opt |= m->mount_opt;
1830 sbi->s_mount_opt &= ~m->mount_opt;
1835 static int parse_options(char *options, struct super_block *sb,
1836 unsigned long *journal_devnum,
1837 unsigned int *journal_ioprio,
1840 struct ext4_sb_info *sbi = EXT4_SB(sb);
1842 substring_t args[MAX_OPT_ARGS];
1848 while ((p = strsep(&options, ",")) != NULL) {
1852 * Initialize args struct so we know whether arg was
1853 * found; some options take optional arguments.
1855 args[0].to = args[0].from = NULL;
1856 token = match_token(p, tokens, args);
1857 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1858 journal_ioprio, is_remount) < 0)
1863 * We do the test below only for project quotas. 'usrquota' and
1864 * 'grpquota' mount options are allowed even without quota feature
1865 * to support legacy quotas in quota files.
1867 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1868 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1869 "Cannot enable project quota enforcement.");
1872 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1873 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1874 clear_opt(sb, USRQUOTA);
1876 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1877 clear_opt(sb, GRPQUOTA);
1879 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1880 ext4_msg(sb, KERN_ERR, "old and new quota "
1885 if (!sbi->s_jquota_fmt) {
1886 ext4_msg(sb, KERN_ERR, "journaled quota format "
1892 if (test_opt(sb, DIOREAD_NOLOCK)) {
1894 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1896 if (blocksize < PAGE_SIZE) {
1897 ext4_msg(sb, KERN_ERR, "can't mount with "
1898 "dioread_nolock if block size != PAGE_SIZE");
1902 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1903 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1904 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1905 "in data=ordered mode");
1911 static inline void ext4_show_quota_options(struct seq_file *seq,
1912 struct super_block *sb)
1914 #if defined(CONFIG_QUOTA)
1915 struct ext4_sb_info *sbi = EXT4_SB(sb);
1917 if (sbi->s_jquota_fmt) {
1920 switch (sbi->s_jquota_fmt) {
1931 seq_printf(seq, ",jqfmt=%s", fmtname);
1934 if (sbi->s_qf_names[USRQUOTA])
1935 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1937 if (sbi->s_qf_names[GRPQUOTA])
1938 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1942 static const char *token2str(int token)
1944 const struct match_token *t;
1946 for (t = tokens; t->token != Opt_err; t++)
1947 if (t->token == token && !strchr(t->pattern, '='))
1954 * - it's set to a non-default value OR
1955 * - if the per-sb default is different from the global default
1957 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1960 struct ext4_sb_info *sbi = EXT4_SB(sb);
1961 struct ext4_super_block *es = sbi->s_es;
1962 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1963 const struct mount_opts *m;
1964 char sep = nodefs ? '\n' : ',';
1966 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1967 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1969 if (sbi->s_sb_block != 1)
1970 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1972 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1973 int want_set = m->flags & MOPT_SET;
1974 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1975 (m->flags & MOPT_CLEAR_ERR))
1977 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1978 continue; /* skip if same as the default */
1980 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1981 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1982 continue; /* select Opt_noFoo vs Opt_Foo */
1983 SEQ_OPTS_PRINT("%s", token2str(m->token));
1986 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1987 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1988 SEQ_OPTS_PRINT("resuid=%u",
1989 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1990 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1991 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1992 SEQ_OPTS_PRINT("resgid=%u",
1993 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1994 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1995 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1996 SEQ_OPTS_PUTS("errors=remount-ro");
1997 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1998 SEQ_OPTS_PUTS("errors=continue");
1999 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2000 SEQ_OPTS_PUTS("errors=panic");
2001 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2002 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2003 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2004 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2005 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2006 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2007 if (sb->s_flags & MS_I_VERSION)
2008 SEQ_OPTS_PUTS("i_version");
2009 if (nodefs || sbi->s_stripe)
2010 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2011 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2012 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2013 SEQ_OPTS_PUTS("data=journal");
2014 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2015 SEQ_OPTS_PUTS("data=ordered");
2016 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2017 SEQ_OPTS_PUTS("data=writeback");
2020 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2021 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2022 sbi->s_inode_readahead_blks);
2024 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2025 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2026 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2027 if (nodefs || sbi->s_max_dir_size_kb)
2028 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2029 if (test_opt(sb, DATA_ERR_ABORT))
2030 SEQ_OPTS_PUTS("data_err=abort");
2031 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2032 SEQ_OPTS_PUTS("test_dummy_encryption");
2034 ext4_show_quota_options(seq, sb);
2038 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2040 return _ext4_show_options(seq, root->d_sb, 0);
2043 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2045 struct super_block *sb = seq->private;
2048 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2049 rc = _ext4_show_options(seq, sb, 1);
2050 seq_puts(seq, "\n");
2054 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2057 struct ext4_sb_info *sbi = EXT4_SB(sb);
2060 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2061 ext4_msg(sb, KERN_ERR, "revision level too high, "
2062 "forcing read-only mode");
2067 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2068 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2069 "running e2fsck is recommended");
2070 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2071 ext4_msg(sb, KERN_WARNING,
2072 "warning: mounting fs with errors, "
2073 "running e2fsck is recommended");
2074 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2075 le16_to_cpu(es->s_mnt_count) >=
2076 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2077 ext4_msg(sb, KERN_WARNING,
2078 "warning: maximal mount count reached, "
2079 "running e2fsck is recommended");
2080 else if (le32_to_cpu(es->s_checkinterval) &&
2081 (le32_to_cpu(es->s_lastcheck) +
2082 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2083 ext4_msg(sb, KERN_WARNING,
2084 "warning: checktime reached, "
2085 "running e2fsck is recommended");
2086 if (!sbi->s_journal)
2087 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2088 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2089 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2090 le16_add_cpu(&es->s_mnt_count, 1);
2091 es->s_mtime = cpu_to_le32(get_seconds());
2092 ext4_update_dynamic_rev(sb);
2094 ext4_set_feature_journal_needs_recovery(sb);
2096 ext4_commit_super(sb, 1);
2098 if (test_opt(sb, DEBUG))
2099 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2100 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2102 sbi->s_groups_count,
2103 EXT4_BLOCKS_PER_GROUP(sb),
2104 EXT4_INODES_PER_GROUP(sb),
2105 sbi->s_mount_opt, sbi->s_mount_opt2);
2107 cleancache_init_fs(sb);
2111 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2113 struct ext4_sb_info *sbi = EXT4_SB(sb);
2114 struct flex_groups **old_groups, **new_groups;
2117 if (!sbi->s_log_groups_per_flex)
2120 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2121 if (size <= sbi->s_flex_groups_allocated)
2124 new_groups = ext4_kvzalloc(roundup_pow_of_two(size *
2125 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2127 ext4_msg(sb, KERN_ERR,
2128 "not enough memory for %d flex group pointers", size);
2131 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2132 new_groups[i] = ext4_kvzalloc(roundup_pow_of_two(
2133 sizeof(struct flex_groups)),
2135 if (!new_groups[i]) {
2136 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2137 kvfree(new_groups[j]);
2139 ext4_msg(sb, KERN_ERR,
2140 "not enough memory for %d flex groups", size);
2145 old_groups = rcu_dereference(sbi->s_flex_groups);
2147 memcpy(new_groups, old_groups,
2148 (sbi->s_flex_groups_allocated *
2149 sizeof(struct flex_groups *)));
2151 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2152 sbi->s_flex_groups_allocated = size;
2154 ext4_kvfree_array_rcu(old_groups);
2158 static int ext4_fill_flex_info(struct super_block *sb)
2160 struct ext4_sb_info *sbi = EXT4_SB(sb);
2161 struct ext4_group_desc *gdp = NULL;
2162 struct flex_groups *fg;
2163 ext4_group_t flex_group;
2166 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2167 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2168 sbi->s_log_groups_per_flex = 0;
2172 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2176 for (i = 0; i < sbi->s_groups_count; i++) {
2177 gdp = ext4_get_group_desc(sb, i, NULL);
2179 flex_group = ext4_flex_group(sbi, i);
2180 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2181 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2182 atomic64_add(ext4_free_group_clusters(sb, gdp),
2183 &fg->free_clusters);
2184 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2192 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2193 struct ext4_group_desc *gdp)
2195 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2197 __le32 le_group = cpu_to_le32(block_group);
2198 struct ext4_sb_info *sbi = EXT4_SB(sb);
2200 if (ext4_has_metadata_csum(sbi->s_sb)) {
2201 /* Use new metadata_csum algorithm */
2203 __u16 dummy_csum = 0;
2205 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2207 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2208 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2209 sizeof(dummy_csum));
2210 offset += sizeof(dummy_csum);
2211 if (offset < sbi->s_desc_size)
2212 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2213 sbi->s_desc_size - offset);
2215 crc = csum32 & 0xFFFF;
2219 /* old crc16 code */
2220 if (!ext4_has_feature_gdt_csum(sb))
2223 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2224 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2225 crc = crc16(crc, (__u8 *)gdp, offset);
2226 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2227 /* for checksum of struct ext4_group_desc do the rest...*/
2228 if (ext4_has_feature_64bit(sb) &&
2229 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2230 crc = crc16(crc, (__u8 *)gdp + offset,
2231 le16_to_cpu(sbi->s_es->s_desc_size) -
2235 return cpu_to_le16(crc);
2238 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2239 struct ext4_group_desc *gdp)
2241 if (ext4_has_group_desc_csum(sb) &&
2242 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2248 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2249 struct ext4_group_desc *gdp)
2251 if (!ext4_has_group_desc_csum(sb))
2253 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2256 /* Called at mount-time, super-block is locked */
2257 static int ext4_check_descriptors(struct super_block *sb,
2258 ext4_fsblk_t sb_block,
2259 ext4_group_t *first_not_zeroed)
2261 struct ext4_sb_info *sbi = EXT4_SB(sb);
2262 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2263 ext4_fsblk_t last_block;
2264 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2265 ext4_fsblk_t block_bitmap;
2266 ext4_fsblk_t inode_bitmap;
2267 ext4_fsblk_t inode_table;
2268 int flexbg_flag = 0;
2269 ext4_group_t i, grp = sbi->s_groups_count;
2271 if (ext4_has_feature_flex_bg(sb))
2274 ext4_debug("Checking group descriptors");
2276 for (i = 0; i < sbi->s_groups_count; i++) {
2277 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2279 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2280 last_block = ext4_blocks_count(sbi->s_es) - 1;
2282 last_block = first_block +
2283 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2285 if ((grp == sbi->s_groups_count) &&
2286 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2289 block_bitmap = ext4_block_bitmap(sb, gdp);
2290 if (block_bitmap == sb_block) {
2291 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2292 "Block bitmap for group %u overlaps "
2294 if (!(sb->s_flags & MS_RDONLY))
2297 if (block_bitmap >= sb_block + 1 &&
2298 block_bitmap <= last_bg_block) {
2299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2300 "Block bitmap for group %u overlaps "
2301 "block group descriptors", i);
2302 if (!(sb->s_flags & MS_RDONLY))
2305 if (block_bitmap < first_block || block_bitmap > last_block) {
2306 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2307 "Block bitmap for group %u not in group "
2308 "(block %llu)!", i, block_bitmap);
2311 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2312 if (inode_bitmap == sb_block) {
2313 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2314 "Inode bitmap for group %u overlaps "
2316 if (!(sb->s_flags & MS_RDONLY))
2319 if (inode_bitmap >= sb_block + 1 &&
2320 inode_bitmap <= last_bg_block) {
2321 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2322 "Inode bitmap for group %u overlaps "
2323 "block group descriptors", i);
2324 if (!(sb->s_flags & MS_RDONLY))
2327 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2328 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2329 "Inode bitmap for group %u not in group "
2330 "(block %llu)!", i, inode_bitmap);
2333 inode_table = ext4_inode_table(sb, gdp);
2334 if (inode_table == sb_block) {
2335 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2336 "Inode table for group %u overlaps "
2338 if (!(sb->s_flags & MS_RDONLY))
2341 if (inode_table >= sb_block + 1 &&
2342 inode_table <= last_bg_block) {
2343 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2344 "Inode table for group %u overlaps "
2345 "block group descriptors", i);
2346 if (!(sb->s_flags & MS_RDONLY))
2349 if (inode_table < first_block ||
2350 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2351 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2352 "Inode table for group %u not in group "
2353 "(block %llu)!", i, inode_table);
2356 ext4_lock_group(sb, i);
2357 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2358 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2359 "Checksum for group %u failed (%u!=%u)",
2360 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2361 gdp)), le16_to_cpu(gdp->bg_checksum));
2362 if (!(sb->s_flags & MS_RDONLY)) {
2363 ext4_unlock_group(sb, i);
2367 ext4_unlock_group(sb, i);
2369 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2371 if (NULL != first_not_zeroed)
2372 *first_not_zeroed = grp;
2376 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2377 * the superblock) which were deleted from all directories, but held open by
2378 * a process at the time of a crash. We walk the list and try to delete these
2379 * inodes at recovery time (only with a read-write filesystem).
2381 * In order to keep the orphan inode chain consistent during traversal (in
2382 * case of crash during recovery), we link each inode into the superblock
2383 * orphan list_head and handle it the same way as an inode deletion during
2384 * normal operation (which journals the operations for us).
2386 * We only do an iget() and an iput() on each inode, which is very safe if we
2387 * accidentally point at an in-use or already deleted inode. The worst that
2388 * can happen in this case is that we get a "bit already cleared" message from
2389 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2390 * e2fsck was run on this filesystem, and it must have already done the orphan
2391 * inode cleanup for us, so we can safely abort without any further action.
2393 static void ext4_orphan_cleanup(struct super_block *sb,
2394 struct ext4_super_block *es)
2396 unsigned int s_flags = sb->s_flags;
2397 int nr_orphans = 0, nr_truncates = 0;
2399 int quota_update = 0;
2402 if (!es->s_last_orphan) {
2403 jbd_debug(4, "no orphan inodes to clean up\n");
2407 if (bdev_read_only(sb->s_bdev)) {
2408 ext4_msg(sb, KERN_ERR, "write access "
2409 "unavailable, skipping orphan cleanup");
2413 /* Check if feature set would not allow a r/w mount */
2414 if (!ext4_feature_set_ok(sb, 0)) {
2415 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2416 "unknown ROCOMPAT features");
2420 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2421 /* don't clear list on RO mount w/ errors */
2422 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2423 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2424 "clearing orphan list.\n");
2425 es->s_last_orphan = 0;
2427 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2431 if (s_flags & MS_RDONLY) {
2432 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2433 sb->s_flags &= ~MS_RDONLY;
2436 /* Needed for iput() to work correctly and not trash data */
2437 sb->s_flags |= MS_ACTIVE;
2440 * Turn on quotas which were not enabled for read-only mounts if
2441 * filesystem has quota feature, so that they are updated correctly.
2443 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2444 int ret = ext4_enable_quotas(sb);
2449 ext4_msg(sb, KERN_ERR,
2450 "Cannot turn on quotas: error %d", ret);
2453 /* Turn on journaled quotas used for old sytle */
2454 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2455 if (EXT4_SB(sb)->s_qf_names[i]) {
2456 int ret = ext4_quota_on_mount(sb, i);
2461 ext4_msg(sb, KERN_ERR,
2462 "Cannot turn on journaled "
2463 "quota: type %d: error %d", i, ret);
2468 while (es->s_last_orphan) {
2469 struct inode *inode;
2472 * We may have encountered an error during cleanup; if
2473 * so, skip the rest.
2475 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2476 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2477 es->s_last_orphan = 0;
2481 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2482 if (IS_ERR(inode)) {
2483 es->s_last_orphan = 0;
2487 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2488 dquot_initialize(inode);
2489 if (inode->i_nlink) {
2490 if (test_opt(sb, DEBUG))
2491 ext4_msg(sb, KERN_DEBUG,
2492 "%s: truncating inode %lu to %lld bytes",
2493 __func__, inode->i_ino, inode->i_size);
2494 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2495 inode->i_ino, inode->i_size);
2497 truncate_inode_pages(inode->i_mapping, inode->i_size);
2498 ext4_truncate(inode);
2499 inode_unlock(inode);
2502 if (test_opt(sb, DEBUG))
2503 ext4_msg(sb, KERN_DEBUG,
2504 "%s: deleting unreferenced inode %lu",
2505 __func__, inode->i_ino);
2506 jbd_debug(2, "deleting unreferenced inode %lu\n",
2510 iput(inode); /* The delete magic happens here! */
2513 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2516 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2517 PLURAL(nr_orphans));
2519 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2520 PLURAL(nr_truncates));
2522 /* Turn off quotas if they were enabled for orphan cleanup */
2524 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2525 if (sb_dqopt(sb)->files[i])
2526 dquot_quota_off(sb, i);
2530 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2534 * Maximal extent format file size.
2535 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2536 * extent format containers, within a sector_t, and within i_blocks
2537 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2538 * so that won't be a limiting factor.
2540 * However there is other limiting factor. We do store extents in the form
2541 * of starting block and length, hence the resulting length of the extent
2542 * covering maximum file size must fit into on-disk format containers as
2543 * well. Given that length is always by 1 unit bigger than max unit (because
2544 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2546 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2548 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2551 loff_t upper_limit = MAX_LFS_FILESIZE;
2553 /* small i_blocks in vfs inode? */
2554 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2556 * CONFIG_LBDAF is not enabled implies the inode
2557 * i_block represent total blocks in 512 bytes
2558 * 32 == size of vfs inode i_blocks * 8
2560 upper_limit = (1LL << 32) - 1;
2562 /* total blocks in file system block size */
2563 upper_limit >>= (blkbits - 9);
2564 upper_limit <<= blkbits;
2568 * 32-bit extent-start container, ee_block. We lower the maxbytes
2569 * by one fs block, so ee_len can cover the extent of maximum file
2572 res = (1LL << 32) - 1;
2575 /* Sanity check against vm- & vfs- imposed limits */
2576 if (res > upper_limit)
2583 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2584 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2585 * We need to be 1 filesystem block less than the 2^48 sector limit.
2587 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2589 loff_t res = EXT4_NDIR_BLOCKS;
2592 /* This is calculated to be the largest file size for a dense, block
2593 * mapped file such that the file's total number of 512-byte sectors,
2594 * including data and all indirect blocks, does not exceed (2^48 - 1).
2596 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2597 * number of 512-byte sectors of the file.
2600 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2602 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2603 * the inode i_block field represents total file blocks in
2604 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2606 upper_limit = (1LL << 32) - 1;
2608 /* total blocks in file system block size */
2609 upper_limit >>= (bits - 9);
2613 * We use 48 bit ext4_inode i_blocks
2614 * With EXT4_HUGE_FILE_FL set the i_blocks
2615 * represent total number of blocks in
2616 * file system block size
2618 upper_limit = (1LL << 48) - 1;
2622 /* indirect blocks */
2624 /* double indirect blocks */
2625 meta_blocks += 1 + (1LL << (bits-2));
2626 /* tripple indirect blocks */
2627 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2629 upper_limit -= meta_blocks;
2630 upper_limit <<= bits;
2632 res += 1LL << (bits-2);
2633 res += 1LL << (2*(bits-2));
2634 res += 1LL << (3*(bits-2));
2636 if (res > upper_limit)
2639 if (res > MAX_LFS_FILESIZE)
2640 res = MAX_LFS_FILESIZE;
2645 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2646 ext4_fsblk_t logical_sb_block, int nr)
2648 struct ext4_sb_info *sbi = EXT4_SB(sb);
2649 ext4_group_t bg, first_meta_bg;
2652 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2654 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2655 return logical_sb_block + nr + 1;
2656 bg = sbi->s_desc_per_block * nr;
2657 if (ext4_bg_has_super(sb, bg))
2661 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2662 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2663 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2666 if (sb->s_blocksize == 1024 && nr == 0 &&
2667 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2670 return (has_super + ext4_group_first_block_no(sb, bg));
2674 * ext4_get_stripe_size: Get the stripe size.
2675 * @sbi: In memory super block info
2677 * If we have specified it via mount option, then
2678 * use the mount option value. If the value specified at mount time is
2679 * greater than the blocks per group use the super block value.
2680 * If the super block value is greater than blocks per group return 0.
2681 * Allocator needs it be less than blocks per group.
2684 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2686 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2687 unsigned long stripe_width =
2688 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2691 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2692 ret = sbi->s_stripe;
2693 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2695 else if (stride && stride <= sbi->s_blocks_per_group)
2701 * If the stripe width is 1, this makes no sense and
2702 * we set it to 0 to turn off stripe handling code.
2711 * Check whether this filesystem can be mounted based on
2712 * the features present and the RDONLY/RDWR mount requested.
2713 * Returns 1 if this filesystem can be mounted as requested,
2714 * 0 if it cannot be.
2716 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2718 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2719 ext4_msg(sb, KERN_ERR,
2720 "Couldn't mount because of "
2721 "unsupported optional features (%x)",
2722 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2723 ~EXT4_FEATURE_INCOMPAT_SUPP));
2730 if (ext4_has_feature_readonly(sb)) {
2731 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2732 sb->s_flags |= MS_RDONLY;
2736 /* Check that feature set is OK for a read-write mount */
2737 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2738 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2739 "unsupported optional features (%x)",
2740 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2741 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2745 * Large file size enabled file system can only be mounted
2746 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2748 if (ext4_has_feature_huge_file(sb)) {
2749 if (sizeof(blkcnt_t) < sizeof(u64)) {
2750 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2751 "cannot be mounted RDWR without "
2756 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2757 ext4_msg(sb, KERN_ERR,
2758 "Can't support bigalloc feature without "
2759 "extents feature\n");
2763 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
2764 if (!readonly && (ext4_has_feature_quota(sb) ||
2765 ext4_has_feature_project(sb))) {
2766 ext4_msg(sb, KERN_ERR,
2767 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
2770 #endif /* CONFIG_QUOTA */
2775 * This function is called once a day if we have errors logged
2776 * on the file system
2778 static void print_daily_error_info(unsigned long arg)
2780 struct super_block *sb = (struct super_block *) arg;
2781 struct ext4_sb_info *sbi;
2782 struct ext4_super_block *es;
2787 if (es->s_error_count)
2788 /* fsck newer than v1.41.13 is needed to clean this condition. */
2789 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2790 le32_to_cpu(es->s_error_count));
2791 if (es->s_first_error_time) {
2792 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2793 sb->s_id, le32_to_cpu(es->s_first_error_time),
2794 (int) sizeof(es->s_first_error_func),
2795 es->s_first_error_func,
2796 le32_to_cpu(es->s_first_error_line));
2797 if (es->s_first_error_ino)
2798 printk(KERN_CONT ": inode %u",
2799 le32_to_cpu(es->s_first_error_ino));
2800 if (es->s_first_error_block)
2801 printk(KERN_CONT ": block %llu", (unsigned long long)
2802 le64_to_cpu(es->s_first_error_block));
2803 printk(KERN_CONT "\n");
2805 if (es->s_last_error_time) {
2806 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2807 sb->s_id, le32_to_cpu(es->s_last_error_time),
2808 (int) sizeof(es->s_last_error_func),
2809 es->s_last_error_func,
2810 le32_to_cpu(es->s_last_error_line));
2811 if (es->s_last_error_ino)
2812 printk(KERN_CONT ": inode %u",
2813 le32_to_cpu(es->s_last_error_ino));
2814 if (es->s_last_error_block)
2815 printk(KERN_CONT ": block %llu", (unsigned long long)
2816 le64_to_cpu(es->s_last_error_block));
2817 printk(KERN_CONT "\n");
2819 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2822 /* Find next suitable group and run ext4_init_inode_table */
2823 static int ext4_run_li_request(struct ext4_li_request *elr)
2825 struct ext4_group_desc *gdp = NULL;
2826 ext4_group_t group, ngroups;
2827 struct super_block *sb;
2828 unsigned long timeout = 0;
2832 ngroups = EXT4_SB(sb)->s_groups_count;
2834 for (group = elr->lr_next_group; group < ngroups; group++) {
2835 gdp = ext4_get_group_desc(sb, group, NULL);
2841 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2845 if (group >= ngroups)
2850 ret = ext4_init_inode_table(sb, group,
2851 elr->lr_timeout ? 0 : 1);
2852 if (elr->lr_timeout == 0) {
2853 timeout = (jiffies - timeout) *
2854 elr->lr_sbi->s_li_wait_mult;
2855 elr->lr_timeout = timeout;
2857 elr->lr_next_sched = jiffies + elr->lr_timeout;
2858 elr->lr_next_group = group + 1;
2864 * Remove lr_request from the list_request and free the
2865 * request structure. Should be called with li_list_mtx held
2867 static void ext4_remove_li_request(struct ext4_li_request *elr)
2869 struct ext4_sb_info *sbi;
2876 list_del(&elr->lr_request);
2877 sbi->s_li_request = NULL;
2881 static void ext4_unregister_li_request(struct super_block *sb)
2883 mutex_lock(&ext4_li_mtx);
2884 if (!ext4_li_info) {
2885 mutex_unlock(&ext4_li_mtx);
2889 mutex_lock(&ext4_li_info->li_list_mtx);
2890 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2891 mutex_unlock(&ext4_li_info->li_list_mtx);
2892 mutex_unlock(&ext4_li_mtx);
2895 static struct task_struct *ext4_lazyinit_task;
2898 * This is the function where ext4lazyinit thread lives. It walks
2899 * through the request list searching for next scheduled filesystem.
2900 * When such a fs is found, run the lazy initialization request
2901 * (ext4_rn_li_request) and keep track of the time spend in this
2902 * function. Based on that time we compute next schedule time of
2903 * the request. When walking through the list is complete, compute
2904 * next waking time and put itself into sleep.
2906 static int ext4_lazyinit_thread(void *arg)
2908 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2909 struct list_head *pos, *n;
2910 struct ext4_li_request *elr;
2911 unsigned long next_wakeup, cur;
2913 BUG_ON(NULL == eli);
2917 next_wakeup = MAX_JIFFY_OFFSET;
2919 mutex_lock(&eli->li_list_mtx);
2920 if (list_empty(&eli->li_request_list)) {
2921 mutex_unlock(&eli->li_list_mtx);
2924 list_for_each_safe(pos, n, &eli->li_request_list) {
2927 elr = list_entry(pos, struct ext4_li_request,
2930 if (time_before(jiffies, elr->lr_next_sched)) {
2931 if (time_before(elr->lr_next_sched, next_wakeup))
2932 next_wakeup = elr->lr_next_sched;
2935 if (down_read_trylock(&elr->lr_super->s_umount)) {
2936 if (sb_start_write_trylock(elr->lr_super)) {
2939 * We hold sb->s_umount, sb can not
2940 * be removed from the list, it is
2941 * now safe to drop li_list_mtx
2943 mutex_unlock(&eli->li_list_mtx);
2944 err = ext4_run_li_request(elr);
2945 sb_end_write(elr->lr_super);
2946 mutex_lock(&eli->li_list_mtx);
2949 up_read((&elr->lr_super->s_umount));
2951 /* error, remove the lazy_init job */
2953 ext4_remove_li_request(elr);
2957 elr->lr_next_sched = jiffies +
2959 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2961 if (time_before(elr->lr_next_sched, next_wakeup))
2962 next_wakeup = elr->lr_next_sched;
2964 mutex_unlock(&eli->li_list_mtx);
2969 if ((time_after_eq(cur, next_wakeup)) ||
2970 (MAX_JIFFY_OFFSET == next_wakeup)) {
2975 schedule_timeout_interruptible(next_wakeup - cur);
2977 if (kthread_should_stop()) {
2978 ext4_clear_request_list();
2985 * It looks like the request list is empty, but we need
2986 * to check it under the li_list_mtx lock, to prevent any
2987 * additions into it, and of course we should lock ext4_li_mtx
2988 * to atomically free the list and ext4_li_info, because at
2989 * this point another ext4 filesystem could be registering
2992 mutex_lock(&ext4_li_mtx);
2993 mutex_lock(&eli->li_list_mtx);
2994 if (!list_empty(&eli->li_request_list)) {
2995 mutex_unlock(&eli->li_list_mtx);
2996 mutex_unlock(&ext4_li_mtx);
2999 mutex_unlock(&eli->li_list_mtx);
3000 kfree(ext4_li_info);
3001 ext4_li_info = NULL;
3002 mutex_unlock(&ext4_li_mtx);
3007 static void ext4_clear_request_list(void)
3009 struct list_head *pos, *n;
3010 struct ext4_li_request *elr;
3012 mutex_lock(&ext4_li_info->li_list_mtx);
3013 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3014 elr = list_entry(pos, struct ext4_li_request,
3016 ext4_remove_li_request(elr);
3018 mutex_unlock(&ext4_li_info->li_list_mtx);
3021 static int ext4_run_lazyinit_thread(void)
3023 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3024 ext4_li_info, "ext4lazyinit");
3025 if (IS_ERR(ext4_lazyinit_task)) {
3026 int err = PTR_ERR(ext4_lazyinit_task);
3027 ext4_clear_request_list();
3028 kfree(ext4_li_info);
3029 ext4_li_info = NULL;
3030 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3031 "initialization thread\n",
3035 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3040 * Check whether it make sense to run itable init. thread or not.
3041 * If there is at least one uninitialized inode table, return
3042 * corresponding group number, else the loop goes through all
3043 * groups and return total number of groups.
3045 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3047 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3048 struct ext4_group_desc *gdp = NULL;
3050 if (!ext4_has_group_desc_csum(sb))
3053 for (group = 0; group < ngroups; group++) {
3054 gdp = ext4_get_group_desc(sb, group, NULL);
3058 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3065 static int ext4_li_info_new(void)
3067 struct ext4_lazy_init *eli = NULL;
3069 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3073 INIT_LIST_HEAD(&eli->li_request_list);
3074 mutex_init(&eli->li_list_mtx);
3076 eli->li_state |= EXT4_LAZYINIT_QUIT;
3083 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3086 struct ext4_sb_info *sbi = EXT4_SB(sb);
3087 struct ext4_li_request *elr;
3089 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3095 elr->lr_next_group = start;
3098 * Randomize first schedule time of the request to
3099 * spread the inode table initialization requests
3102 elr->lr_next_sched = jiffies + (prandom_u32() %
3103 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3107 int ext4_register_li_request(struct super_block *sb,
3108 ext4_group_t first_not_zeroed)
3110 struct ext4_sb_info *sbi = EXT4_SB(sb);
3111 struct ext4_li_request *elr = NULL;
3112 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3115 mutex_lock(&ext4_li_mtx);
3116 if (sbi->s_li_request != NULL) {
3118 * Reset timeout so it can be computed again, because
3119 * s_li_wait_mult might have changed.
3121 sbi->s_li_request->lr_timeout = 0;
3125 if (first_not_zeroed == ngroups ||
3126 (sb->s_flags & MS_RDONLY) ||
3127 !test_opt(sb, INIT_INODE_TABLE))
3130 elr = ext4_li_request_new(sb, first_not_zeroed);
3136 if (NULL == ext4_li_info) {
3137 ret = ext4_li_info_new();
3142 mutex_lock(&ext4_li_info->li_list_mtx);
3143 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3144 mutex_unlock(&ext4_li_info->li_list_mtx);
3146 sbi->s_li_request = elr;
3148 * set elr to NULL here since it has been inserted to
3149 * the request_list and the removal and free of it is
3150 * handled by ext4_clear_request_list from now on.
3154 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3155 ret = ext4_run_lazyinit_thread();
3160 mutex_unlock(&ext4_li_mtx);
3167 * We do not need to lock anything since this is called on
3170 static void ext4_destroy_lazyinit_thread(void)
3173 * If thread exited earlier
3174 * there's nothing to be done.
3176 if (!ext4_li_info || !ext4_lazyinit_task)
3179 kthread_stop(ext4_lazyinit_task);
3182 static int set_journal_csum_feature_set(struct super_block *sb)
3185 int compat, incompat;
3186 struct ext4_sb_info *sbi = EXT4_SB(sb);
3188 if (ext4_has_metadata_csum(sb)) {
3189 /* journal checksum v3 */
3191 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3193 /* journal checksum v1 */
3194 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3198 jbd2_journal_clear_features(sbi->s_journal,
3199 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3200 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3201 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3202 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3203 ret = jbd2_journal_set_features(sbi->s_journal,
3205 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3207 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3208 ret = jbd2_journal_set_features(sbi->s_journal,
3211 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3212 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3214 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3215 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3222 * Note: calculating the overhead so we can be compatible with
3223 * historical BSD practice is quite difficult in the face of
3224 * clusters/bigalloc. This is because multiple metadata blocks from
3225 * different block group can end up in the same allocation cluster.
3226 * Calculating the exact overhead in the face of clustered allocation
3227 * requires either O(all block bitmaps) in memory or O(number of block
3228 * groups**2) in time. We will still calculate the superblock for
3229 * older file systems --- and if we come across with a bigalloc file
3230 * system with zero in s_overhead_clusters the estimate will be close to
3231 * correct especially for very large cluster sizes --- but for newer
3232 * file systems, it's better to calculate this figure once at mkfs
3233 * time, and store it in the superblock. If the superblock value is
3234 * present (even for non-bigalloc file systems), we will use it.
3236 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3239 struct ext4_sb_info *sbi = EXT4_SB(sb);
3240 struct ext4_group_desc *gdp;
3241 ext4_fsblk_t first_block, last_block, b;
3242 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3243 int s, j, count = 0;
3245 if (!ext4_has_feature_bigalloc(sb))
3246 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3247 sbi->s_itb_per_group + 2);
3249 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3250 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3251 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3252 for (i = 0; i < ngroups; i++) {
3253 gdp = ext4_get_group_desc(sb, i, NULL);
3254 b = ext4_block_bitmap(sb, gdp);
3255 if (b >= first_block && b <= last_block) {
3256 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3259 b = ext4_inode_bitmap(sb, gdp);
3260 if (b >= first_block && b <= last_block) {
3261 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3264 b = ext4_inode_table(sb, gdp);
3265 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3266 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3267 int c = EXT4_B2C(sbi, b - first_block);
3268 ext4_set_bit(c, buf);
3274 if (ext4_bg_has_super(sb, grp)) {
3275 ext4_set_bit(s++, buf);
3278 j = ext4_bg_num_gdb(sb, grp);
3279 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3280 ext4_error(sb, "Invalid number of block group "
3281 "descriptor blocks: %d", j);
3282 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3286 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3290 return EXT4_CLUSTERS_PER_GROUP(sb) -
3291 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3295 * Compute the overhead and stash it in sbi->s_overhead
3297 int ext4_calculate_overhead(struct super_block *sb)
3299 struct ext4_sb_info *sbi = EXT4_SB(sb);
3300 struct ext4_super_block *es = sbi->s_es;
3301 struct inode *j_inode;
3302 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3303 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3304 ext4_fsblk_t overhead = 0;
3305 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3311 * Compute the overhead (FS structures). This is constant
3312 * for a given filesystem unless the number of block groups
3313 * changes so we cache the previous value until it does.
3317 * All of the blocks before first_data_block are overhead
3319 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3322 * Add the overhead found in each block group
3324 for (i = 0; i < ngroups; i++) {
3327 blks = count_overhead(sb, i, buf);
3330 memset(buf, 0, PAGE_SIZE);
3335 * Add the internal journal blocks whether the journal has been
3338 if (sbi->s_journal && !sbi->journal_bdev)
3339 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3340 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3341 /* j_inum for internal journal is non-zero */
3342 j_inode = ext4_get_journal_inode(sb, j_inum);
3344 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3345 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3348 ext4_msg(sb, KERN_ERR, "can't get journal size");
3351 sbi->s_overhead = overhead;
3353 free_page((unsigned long) buf);
3357 static void ext4_clamp_want_extra_isize(struct super_block *sb)
3359 struct ext4_sb_info *sbi = EXT4_SB(sb);
3360 struct ext4_super_block *es = sbi->s_es;
3361 unsigned def_extra_isize = sizeof(struct ext4_inode) -
3362 EXT4_GOOD_OLD_INODE_SIZE;
3364 if (sbi->s_inode_size == EXT4_GOOD_OLD_INODE_SIZE) {
3365 sbi->s_want_extra_isize = 0;
3368 if (sbi->s_want_extra_isize < 4) {
3369 sbi->s_want_extra_isize = def_extra_isize;
3370 if (ext4_has_feature_extra_isize(sb)) {
3371 if (sbi->s_want_extra_isize <
3372 le16_to_cpu(es->s_want_extra_isize))
3373 sbi->s_want_extra_isize =
3374 le16_to_cpu(es->s_want_extra_isize);
3375 if (sbi->s_want_extra_isize <
3376 le16_to_cpu(es->s_min_extra_isize))
3377 sbi->s_want_extra_isize =
3378 le16_to_cpu(es->s_min_extra_isize);
3381 /* Check if enough inode space is available */
3382 if ((sbi->s_want_extra_isize > sbi->s_inode_size) ||
3383 (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3384 sbi->s_inode_size)) {
3385 sbi->s_want_extra_isize = def_extra_isize;
3386 ext4_msg(sb, KERN_INFO,
3387 "required extra inode space not available");
3391 static void ext4_set_resv_clusters(struct super_block *sb)
3393 ext4_fsblk_t resv_clusters;
3394 struct ext4_sb_info *sbi = EXT4_SB(sb);
3397 * There's no need to reserve anything when we aren't using extents.
3398 * The space estimates are exact, there are no unwritten extents,
3399 * hole punching doesn't need new metadata... This is needed especially
3400 * to keep ext2/3 backward compatibility.
3402 if (!ext4_has_feature_extents(sb))
3405 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3406 * This should cover the situations where we can not afford to run
3407 * out of space like for example punch hole, or converting
3408 * unwritten extents in delalloc path. In most cases such
3409 * allocation would require 1, or 2 blocks, higher numbers are
3412 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3413 sbi->s_cluster_bits);
3415 do_div(resv_clusters, 50);
3416 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3418 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3421 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3423 char *orig_data = kstrdup(data, GFP_KERNEL);
3424 struct buffer_head *bh, **group_desc;
3425 struct ext4_super_block *es = NULL;
3426 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3427 struct flex_groups **flex_groups;
3429 ext4_fsblk_t sb_block = get_sb_block(&data);
3430 ext4_fsblk_t logical_sb_block;
3431 unsigned long offset = 0;
3432 unsigned long journal_devnum = 0;
3433 unsigned long def_mount_opts;
3437 int blocksize, clustersize;
3438 unsigned int db_count;
3440 int needs_recovery, has_huge_files, has_bigalloc;
3443 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3444 ext4_group_t first_not_zeroed;
3446 if ((data && !orig_data) || !sbi)
3449 sbi->s_blockgroup_lock =
3450 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3451 if (!sbi->s_blockgroup_lock)
3454 sb->s_fs_info = sbi;
3456 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3457 sbi->s_sb_block = sb_block;
3458 if (sb->s_bdev->bd_part)
3459 sbi->s_sectors_written_start =
3460 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3462 /* Cleanup superblock name */
3463 strreplace(sb->s_id, '/', '!');
3465 /* -EINVAL is default */
3467 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3469 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3474 * The ext4 superblock will not be buffer aligned for other than 1kB
3475 * block sizes. We need to calculate the offset from buffer start.
3477 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3478 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3479 offset = do_div(logical_sb_block, blocksize);
3481 logical_sb_block = sb_block;
3484 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3485 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3489 * Note: s_es must be initialized as soon as possible because
3490 * some ext4 macro-instructions depend on its value
3492 es = (struct ext4_super_block *) (bh->b_data + offset);
3494 sb->s_magic = le16_to_cpu(es->s_magic);
3495 if (sb->s_magic != EXT4_SUPER_MAGIC)
3497 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3499 /* Warn if metadata_csum and gdt_csum are both set. */
3500 if (ext4_has_feature_metadata_csum(sb) &&
3501 ext4_has_feature_gdt_csum(sb))
3502 ext4_warning(sb, "metadata_csum and uninit_bg are "
3503 "redundant flags; please run fsck.");
3505 /* Check for a known checksum algorithm */
3506 if (!ext4_verify_csum_type(sb, es)) {
3507 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3508 "unknown checksum algorithm.");
3513 /* Load the checksum driver */
3514 if (ext4_has_feature_metadata_csum(sb)) {
3515 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3516 if (IS_ERR(sbi->s_chksum_driver)) {
3517 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3518 ret = PTR_ERR(sbi->s_chksum_driver);
3519 sbi->s_chksum_driver = NULL;
3524 /* Check superblock checksum */
3525 if (!ext4_superblock_csum_verify(sb, es)) {
3526 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3527 "invalid superblock checksum. Run e2fsck?");
3533 /* Precompute checksum seed for all metadata */
3534 if (ext4_has_feature_csum_seed(sb))
3535 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3536 else if (ext4_has_metadata_csum(sb))
3537 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3538 sizeof(es->s_uuid));
3540 /* Set defaults before we parse the mount options */
3541 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3542 set_opt(sb, INIT_INODE_TABLE);
3543 if (def_mount_opts & EXT4_DEFM_DEBUG)
3545 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3547 if (def_mount_opts & EXT4_DEFM_UID16)
3548 set_opt(sb, NO_UID32);
3549 /* xattr user namespace & acls are now defaulted on */
3550 set_opt(sb, XATTR_USER);
3551 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3552 set_opt(sb, POSIX_ACL);
3554 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3555 if (ext4_has_metadata_csum(sb))
3556 set_opt(sb, JOURNAL_CHECKSUM);
3558 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3559 set_opt(sb, JOURNAL_DATA);
3560 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3561 set_opt(sb, ORDERED_DATA);
3562 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3563 set_opt(sb, WRITEBACK_DATA);
3565 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3566 set_opt(sb, ERRORS_PANIC);
3567 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3568 set_opt(sb, ERRORS_CONT);
3570 set_opt(sb, ERRORS_RO);
3571 /* block_validity enabled by default; disable with noblock_validity */
3572 set_opt(sb, BLOCK_VALIDITY);
3573 if (def_mount_opts & EXT4_DEFM_DISCARD)
3574 set_opt(sb, DISCARD);
3576 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3577 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3578 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3579 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3580 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3582 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3583 set_opt(sb, BARRIER);
3586 * enable delayed allocation by default
3587 * Use -o nodelalloc to turn it off
3589 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3590 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3591 set_opt(sb, DELALLOC);
3594 * set default s_li_wait_mult for lazyinit, for the case there is
3595 * no mount option specified.
3597 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3599 if (sbi->s_es->s_mount_opts[0]) {
3600 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3601 sizeof(sbi->s_es->s_mount_opts),
3605 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3606 &journal_ioprio, 0)) {
3607 ext4_msg(sb, KERN_WARNING,
3608 "failed to parse options in superblock: %s",
3611 kfree(s_mount_opts);
3613 sbi->s_def_mount_opt = sbi->s_mount_opt;
3614 if (!parse_options((char *) data, sb, &journal_devnum,
3615 &journal_ioprio, 0))
3618 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3619 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3620 "with data=journal disables delayed "
3621 "allocation and O_DIRECT support!\n");
3622 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3623 ext4_msg(sb, KERN_ERR, "can't mount with "
3624 "both data=journal and delalloc");
3627 if (test_opt(sb, DIOREAD_NOLOCK)) {
3628 ext4_msg(sb, KERN_ERR, "can't mount with "
3629 "both data=journal and dioread_nolock");
3632 if (test_opt(sb, DAX)) {
3633 ext4_msg(sb, KERN_ERR, "can't mount with "
3634 "both data=journal and dax");
3637 if (ext4_has_feature_encrypt(sb)) {
3638 ext4_msg(sb, KERN_WARNING,
3639 "encrypted files will use data=ordered "
3640 "instead of data journaling mode");
3642 if (test_opt(sb, DELALLOC))
3643 clear_opt(sb, DELALLOC);
3645 sb->s_iflags |= SB_I_CGROUPWB;
3648 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3649 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3651 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3652 (ext4_has_compat_features(sb) ||
3653 ext4_has_ro_compat_features(sb) ||
3654 ext4_has_incompat_features(sb)))
3655 ext4_msg(sb, KERN_WARNING,
3656 "feature flags set on rev 0 fs, "
3657 "running e2fsck is recommended");
3659 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3660 set_opt2(sb, HURD_COMPAT);
3661 if (ext4_has_feature_64bit(sb)) {
3662 ext4_msg(sb, KERN_ERR,
3663 "The Hurd can't support 64-bit file systems");
3668 if (IS_EXT2_SB(sb)) {
3669 if (ext2_feature_set_ok(sb))
3670 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3671 "using the ext4 subsystem");
3673 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3674 "to feature incompatibilities");
3679 if (IS_EXT3_SB(sb)) {
3680 if (ext3_feature_set_ok(sb))
3681 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3682 "using the ext4 subsystem");
3684 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3685 "to feature incompatibilities");
3691 * Check feature flags regardless of the revision level, since we
3692 * previously didn't change the revision level when setting the flags,
3693 * so there is a chance incompat flags are set on a rev 0 filesystem.
3695 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3698 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3699 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3700 blocksize > EXT4_MAX_BLOCK_SIZE) {
3701 ext4_msg(sb, KERN_ERR,
3702 "Unsupported filesystem blocksize %d (%d log_block_size)",
3703 blocksize, le32_to_cpu(es->s_log_block_size));
3706 if (le32_to_cpu(es->s_log_block_size) >
3707 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3708 ext4_msg(sb, KERN_ERR,
3709 "Invalid log block size: %u",
3710 le32_to_cpu(es->s_log_block_size));
3713 if (le32_to_cpu(es->s_log_cluster_size) >
3714 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3715 ext4_msg(sb, KERN_ERR,
3716 "Invalid log cluster size: %u",
3717 le32_to_cpu(es->s_log_cluster_size));
3721 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3722 ext4_msg(sb, KERN_ERR,
3723 "Number of reserved GDT blocks insanely large: %d",
3724 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3728 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3729 err = bdev_dax_supported(sb, blocksize);
3734 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3735 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3736 es->s_encryption_level);
3740 if (sb->s_blocksize != blocksize) {
3741 /* Validate the filesystem blocksize */
3742 if (!sb_set_blocksize(sb, blocksize)) {
3743 ext4_msg(sb, KERN_ERR, "bad block size %d",
3749 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3750 offset = do_div(logical_sb_block, blocksize);
3751 bh = sb_bread_unmovable(sb, logical_sb_block);
3753 ext4_msg(sb, KERN_ERR,
3754 "Can't read superblock on 2nd try");
3757 es = (struct ext4_super_block *)(bh->b_data + offset);
3759 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3760 ext4_msg(sb, KERN_ERR,
3761 "Magic mismatch, very weird!");
3766 has_huge_files = ext4_has_feature_huge_file(sb);
3767 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3769 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3771 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3772 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3773 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3775 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3776 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3777 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3778 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3782 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3783 (!is_power_of_2(sbi->s_inode_size)) ||
3784 (sbi->s_inode_size > blocksize)) {
3785 ext4_msg(sb, KERN_ERR,
3786 "unsupported inode size: %d",
3790 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3791 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3794 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3795 if (ext4_has_feature_64bit(sb)) {
3796 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3797 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3798 !is_power_of_2(sbi->s_desc_size)) {
3799 ext4_msg(sb, KERN_ERR,
3800 "unsupported descriptor size %lu",
3805 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3807 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3808 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3810 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3811 if (sbi->s_inodes_per_block == 0)
3813 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3814 sbi->s_inodes_per_group > blocksize * 8) {
3815 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3816 sbi->s_inodes_per_group);
3819 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3820 sbi->s_inodes_per_block;
3821 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3823 sbi->s_mount_state = le16_to_cpu(es->s_state);
3824 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3825 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3827 for (i = 0; i < 4; i++)
3828 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3829 sbi->s_def_hash_version = es->s_def_hash_version;
3830 if (ext4_has_feature_dir_index(sb)) {
3831 i = le32_to_cpu(es->s_flags);
3832 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3833 sbi->s_hash_unsigned = 3;
3834 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3835 #ifdef __CHAR_UNSIGNED__
3836 if (!(sb->s_flags & MS_RDONLY))
3838 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3839 sbi->s_hash_unsigned = 3;
3841 if (!(sb->s_flags & MS_RDONLY))
3843 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3848 /* Handle clustersize */
3849 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3850 has_bigalloc = ext4_has_feature_bigalloc(sb);
3852 if (clustersize < blocksize) {
3853 ext4_msg(sb, KERN_ERR,
3854 "cluster size (%d) smaller than "
3855 "block size (%d)", clustersize, blocksize);
3858 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3859 le32_to_cpu(es->s_log_block_size);
3860 sbi->s_clusters_per_group =
3861 le32_to_cpu(es->s_clusters_per_group);
3862 if (sbi->s_clusters_per_group > blocksize * 8) {
3863 ext4_msg(sb, KERN_ERR,
3864 "#clusters per group too big: %lu",
3865 sbi->s_clusters_per_group);
3868 if (sbi->s_blocks_per_group !=
3869 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3870 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3871 "clusters per group (%lu) inconsistent",
3872 sbi->s_blocks_per_group,
3873 sbi->s_clusters_per_group);
3877 if (clustersize != blocksize) {
3878 ext4_msg(sb, KERN_ERR,
3879 "fragment/cluster size (%d) != "
3880 "block size (%d)", clustersize, blocksize);
3883 if (sbi->s_blocks_per_group > blocksize * 8) {
3884 ext4_msg(sb, KERN_ERR,
3885 "#blocks per group too big: %lu",
3886 sbi->s_blocks_per_group);
3889 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3890 sbi->s_cluster_bits = 0;
3892 sbi->s_cluster_ratio = clustersize / blocksize;
3894 /* Do we have standard group size of clustersize * 8 blocks ? */
3895 if (sbi->s_blocks_per_group == clustersize << 3)
3896 set_opt2(sb, STD_GROUP_SIZE);
3899 * Test whether we have more sectors than will fit in sector_t,
3900 * and whether the max offset is addressable by the page cache.
3902 err = generic_check_addressable(sb->s_blocksize_bits,
3903 ext4_blocks_count(es));
3905 ext4_msg(sb, KERN_ERR, "filesystem"
3906 " too large to mount safely on this system");
3907 if (sizeof(sector_t) < 8)
3908 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3912 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3915 /* check blocks count against device size */
3916 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3917 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3918 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3919 "exceeds size of device (%llu blocks)",
3920 ext4_blocks_count(es), blocks_count);
3925 * It makes no sense for the first data block to be beyond the end
3926 * of the filesystem.
3928 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3929 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3930 "block %u is beyond end of filesystem (%llu)",
3931 le32_to_cpu(es->s_first_data_block),
3932 ext4_blocks_count(es));
3935 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
3936 (sbi->s_cluster_ratio == 1)) {
3937 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3938 "block is 0 with a 1k block and cluster size");
3942 blocks_count = (ext4_blocks_count(es) -
3943 le32_to_cpu(es->s_first_data_block) +
3944 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3945 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3946 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3947 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
3948 "(block count %llu, first data block %u, "
3949 "blocks per group %lu)", blocks_count,
3950 ext4_blocks_count(es),
3951 le32_to_cpu(es->s_first_data_block),
3952 EXT4_BLOCKS_PER_GROUP(sb));
3955 sbi->s_groups_count = blocks_count;
3956 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3957 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3958 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
3959 le32_to_cpu(es->s_inodes_count)) {
3960 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
3961 le32_to_cpu(es->s_inodes_count),
3962 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
3966 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3967 EXT4_DESC_PER_BLOCK(sb);
3968 if (ext4_has_feature_meta_bg(sb)) {
3969 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3970 ext4_msg(sb, KERN_WARNING,
3971 "first meta block group too large: %u "
3972 "(group descriptor block count %u)",
3973 le32_to_cpu(es->s_first_meta_bg), db_count);
3977 rcu_assign_pointer(sbi->s_group_desc,
3978 ext4_kvmalloc(db_count *
3979 sizeof(struct buffer_head *),
3981 if (sbi->s_group_desc == NULL) {
3982 ext4_msg(sb, KERN_ERR, "not enough memory");
3987 bgl_lock_init(sbi->s_blockgroup_lock);
3989 for (i = 0; i < db_count; i++) {
3990 struct buffer_head *bh;
3992 block = descriptor_loc(sb, logical_sb_block, i);
3993 bh = sb_bread_unmovable(sb, block);
3995 ext4_msg(sb, KERN_ERR,
3996 "can't read group descriptor %d", i);
4001 rcu_dereference(sbi->s_group_desc)[i] = bh;
4004 sbi->s_gdb_count = db_count;
4005 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4006 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4007 ret = -EFSCORRUPTED;
4011 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
4012 spin_lock_init(&sbi->s_next_gen_lock);
4014 setup_timer(&sbi->s_err_report, print_daily_error_info,
4015 (unsigned long) sb);
4017 /* Register extent status tree shrinker */
4018 if (ext4_es_register_shrinker(sbi))
4021 sbi->s_stripe = ext4_get_stripe_size(sbi);
4022 sbi->s_extent_max_zeroout_kb = 32;
4025 * set up enough so that it can read an inode
4027 sb->s_op = &ext4_sops;
4028 sb->s_export_op = &ext4_export_ops;
4029 sb->s_xattr = ext4_xattr_handlers;
4030 sb->s_cop = &ext4_cryptops;
4032 sb->dq_op = &ext4_quota_operations;
4033 if (ext4_has_feature_quota(sb))
4034 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4036 sb->s_qcop = &ext4_qctl_operations;
4037 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4039 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4041 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4042 mutex_init(&sbi->s_orphan_lock);
4046 needs_recovery = (es->s_last_orphan != 0 ||
4047 ext4_has_feature_journal_needs_recovery(sb));
4049 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
4050 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4051 goto failed_mount3a;
4054 * The first inode we look at is the journal inode. Don't try
4055 * root first: it may be modified in the journal!
4057 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4058 err = ext4_load_journal(sb, es, journal_devnum);
4060 goto failed_mount3a;
4061 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4062 ext4_has_feature_journal_needs_recovery(sb)) {
4063 ext4_msg(sb, KERN_ERR, "required journal recovery "
4064 "suppressed and not mounted read-only");
4065 goto failed_mount_wq;
4067 /* Nojournal mode, all journal mount options are illegal */
4068 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4069 ext4_msg(sb, KERN_ERR, "can't mount with "
4070 "journal_checksum, fs mounted w/o journal");
4071 goto failed_mount_wq;
4073 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4074 ext4_msg(sb, KERN_ERR, "can't mount with "
4075 "journal_async_commit, fs mounted w/o journal");
4076 goto failed_mount_wq;
4078 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4079 ext4_msg(sb, KERN_ERR, "can't mount with "
4080 "commit=%lu, fs mounted w/o journal",
4081 sbi->s_commit_interval / HZ);
4082 goto failed_mount_wq;
4084 if (EXT4_MOUNT_DATA_FLAGS &
4085 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4086 ext4_msg(sb, KERN_ERR, "can't mount with "
4087 "data=, fs mounted w/o journal");
4088 goto failed_mount_wq;
4090 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4091 clear_opt(sb, JOURNAL_CHECKSUM);
4092 clear_opt(sb, DATA_FLAGS);
4093 sbi->s_journal = NULL;
4098 if (ext4_has_feature_64bit(sb) &&
4099 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4100 JBD2_FEATURE_INCOMPAT_64BIT)) {
4101 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4102 goto failed_mount_wq;
4105 if (!set_journal_csum_feature_set(sb)) {
4106 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4108 goto failed_mount_wq;
4111 /* We have now updated the journal if required, so we can
4112 * validate the data journaling mode. */
4113 switch (test_opt(sb, DATA_FLAGS)) {
4115 /* No mode set, assume a default based on the journal
4116 * capabilities: ORDERED_DATA if the journal can
4117 * cope, else JOURNAL_DATA
4119 if (jbd2_journal_check_available_features
4120 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4121 set_opt(sb, ORDERED_DATA);
4123 set_opt(sb, JOURNAL_DATA);
4126 case EXT4_MOUNT_ORDERED_DATA:
4127 case EXT4_MOUNT_WRITEBACK_DATA:
4128 if (!jbd2_journal_check_available_features
4129 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4130 ext4_msg(sb, KERN_ERR, "Journal does not support "
4131 "requested data journaling mode");
4132 goto failed_mount_wq;
4137 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4139 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4142 sbi->s_mb_cache = ext4_xattr_create_cache();
4143 if (!sbi->s_mb_cache) {
4144 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4145 goto failed_mount_wq;
4148 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4149 (blocksize != PAGE_SIZE)) {
4150 ext4_msg(sb, KERN_ERR,
4151 "Unsupported blocksize for fs encryption");
4152 goto failed_mount_wq;
4155 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4156 !ext4_has_feature_encrypt(sb)) {
4157 ext4_set_feature_encrypt(sb);
4158 ext4_commit_super(sb, 1);
4162 * Get the # of file system overhead blocks from the
4163 * superblock if present.
4165 if (es->s_overhead_clusters)
4166 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4168 err = ext4_calculate_overhead(sb);
4170 goto failed_mount_wq;
4174 * The maximum number of concurrent works can be high and
4175 * concurrency isn't really necessary. Limit it to 1.
4177 EXT4_SB(sb)->rsv_conversion_wq =
4178 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4179 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4180 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4186 * The jbd2_journal_load will have done any necessary log recovery,
4187 * so we can safely mount the rest of the filesystem now.
4190 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4192 ext4_msg(sb, KERN_ERR, "get root inode failed");
4193 ret = PTR_ERR(root);
4197 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4198 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4202 sb->s_root = d_make_root(root);
4204 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4209 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4210 sb->s_flags |= MS_RDONLY;
4212 ext4_clamp_want_extra_isize(sb);
4214 ext4_set_resv_clusters(sb);
4216 err = ext4_setup_system_zone(sb);
4218 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4220 goto failed_mount4a;
4224 err = ext4_mb_init(sb);
4226 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4231 block = ext4_count_free_clusters(sb);
4232 ext4_free_blocks_count_set(sbi->s_es,
4233 EXT4_C2B(sbi, block));
4234 ext4_superblock_csum_set(sb);
4235 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4238 unsigned long freei = ext4_count_free_inodes(sb);
4239 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4240 ext4_superblock_csum_set(sb);
4241 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4245 err = percpu_counter_init(&sbi->s_dirs_counter,
4246 ext4_count_dirs(sb), GFP_KERNEL);
4248 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4251 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4254 ext4_msg(sb, KERN_ERR, "insufficient memory");
4258 if (ext4_has_feature_flex_bg(sb))
4259 if (!ext4_fill_flex_info(sb)) {
4260 ext4_msg(sb, KERN_ERR,
4261 "unable to initialize "
4262 "flex_bg meta info!");
4266 err = ext4_register_li_request(sb, first_not_zeroed);
4270 err = ext4_register_sysfs(sb);
4275 /* Enable quota usage during mount. */
4276 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4277 err = ext4_enable_quotas(sb);
4281 #endif /* CONFIG_QUOTA */
4283 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4284 ext4_orphan_cleanup(sb, es);
4285 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4286 if (needs_recovery) {
4287 ext4_msg(sb, KERN_INFO, "recovery complete");
4288 ext4_mark_recovery_complete(sb, es);
4290 if (EXT4_SB(sb)->s_journal) {
4291 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4292 descr = " journalled data mode";
4293 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4294 descr = " ordered data mode";
4296 descr = " writeback data mode";
4298 descr = "out journal";
4300 if (test_opt(sb, DISCARD)) {
4301 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4302 if (!blk_queue_discard(q))
4303 ext4_msg(sb, KERN_WARNING,
4304 "mounting with \"discard\" option, but "
4305 "the device does not support discard");
4308 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4309 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4310 "Opts: %.*s%s%s", descr,
4311 (int) sizeof(sbi->s_es->s_mount_opts),
4312 sbi->s_es->s_mount_opts,
4313 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4315 if (es->s_error_count)
4316 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4318 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4319 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4320 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4321 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4324 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4325 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4326 EXT4_KEY_DESC_PREFIX_SIZE);
4327 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4333 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4338 ext4_unregister_sysfs(sb);
4339 kobject_put(&sbi->s_kobj);
4342 ext4_unregister_li_request(sb);
4344 ext4_mb_release(sb);
4346 flex_groups = rcu_dereference(sbi->s_flex_groups);
4348 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4349 kvfree(flex_groups[i]);
4350 kvfree(flex_groups);
4353 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4354 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4355 percpu_counter_destroy(&sbi->s_dirs_counter);
4356 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4357 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4359 ext4_ext_release(sb);
4360 ext4_release_system_zone(sb);
4365 ext4_msg(sb, KERN_ERR, "mount failed");
4366 if (EXT4_SB(sb)->rsv_conversion_wq)
4367 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4369 if (sbi->s_mb_cache) {
4370 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4371 sbi->s_mb_cache = NULL;
4373 if (sbi->s_journal) {
4374 jbd2_journal_destroy(sbi->s_journal);
4375 sbi->s_journal = NULL;
4378 ext4_es_unregister_shrinker(sbi);
4380 del_timer_sync(&sbi->s_err_report);
4382 kthread_stop(sbi->s_mmp_tsk);
4385 group_desc = rcu_dereference(sbi->s_group_desc);
4386 for (i = 0; i < db_count; i++)
4387 brelse(group_desc[i]);
4391 if (sbi->s_chksum_driver)
4392 crypto_free_shash(sbi->s_chksum_driver);
4394 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4395 kfree(sbi->s_qf_names[i]);
4397 ext4_blkdev_remove(sbi);
4400 sb->s_fs_info = NULL;
4401 kfree(sbi->s_blockgroup_lock);
4405 return err ? err : ret;
4409 * Setup any per-fs journal parameters now. We'll do this both on
4410 * initial mount, once the journal has been initialised but before we've
4411 * done any recovery; and again on any subsequent remount.
4413 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4415 struct ext4_sb_info *sbi = EXT4_SB(sb);
4417 journal->j_commit_interval = sbi->s_commit_interval;
4418 journal->j_min_batch_time = sbi->s_min_batch_time;
4419 journal->j_max_batch_time = sbi->s_max_batch_time;
4421 write_lock(&journal->j_state_lock);
4422 if (test_opt(sb, BARRIER))
4423 journal->j_flags |= JBD2_BARRIER;
4425 journal->j_flags &= ~JBD2_BARRIER;
4426 if (test_opt(sb, DATA_ERR_ABORT))
4427 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4429 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4430 write_unlock(&journal->j_state_lock);
4433 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4434 unsigned int journal_inum)
4436 struct inode *journal_inode;
4439 * Test for the existence of a valid inode on disk. Bad things
4440 * happen if we iget() an unused inode, as the subsequent iput()
4441 * will try to delete it.
4443 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4444 if (IS_ERR(journal_inode)) {
4445 ext4_msg(sb, KERN_ERR, "no journal found");
4448 if (!journal_inode->i_nlink) {
4449 make_bad_inode(journal_inode);
4450 iput(journal_inode);
4451 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4455 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4456 journal_inode, journal_inode->i_size);
4457 if (!S_ISREG(journal_inode->i_mode)) {
4458 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4459 iput(journal_inode);
4462 return journal_inode;
4465 static journal_t *ext4_get_journal(struct super_block *sb,
4466 unsigned int journal_inum)
4468 struct inode *journal_inode;
4471 BUG_ON(!ext4_has_feature_journal(sb));
4473 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4477 journal = jbd2_journal_init_inode(journal_inode);
4479 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4480 iput(journal_inode);
4483 journal->j_private = sb;
4484 ext4_init_journal_params(sb, journal);
4488 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4491 struct buffer_head *bh;
4495 int hblock, blocksize;
4496 ext4_fsblk_t sb_block;
4497 unsigned long offset;
4498 struct ext4_super_block *es;
4499 struct block_device *bdev;
4501 BUG_ON(!ext4_has_feature_journal(sb));
4503 bdev = ext4_blkdev_get(j_dev, sb);
4507 blocksize = sb->s_blocksize;
4508 hblock = bdev_logical_block_size(bdev);
4509 if (blocksize < hblock) {
4510 ext4_msg(sb, KERN_ERR,
4511 "blocksize too small for journal device");
4515 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4516 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4517 set_blocksize(bdev, blocksize);
4518 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4519 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4520 "external journal");
4524 es = (struct ext4_super_block *) (bh->b_data + offset);
4525 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4526 !(le32_to_cpu(es->s_feature_incompat) &
4527 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4528 ext4_msg(sb, KERN_ERR, "external journal has "
4534 if ((le32_to_cpu(es->s_feature_ro_compat) &
4535 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4536 es->s_checksum != ext4_superblock_csum(sb, es)) {
4537 ext4_msg(sb, KERN_ERR, "external journal has "
4538 "corrupt superblock");
4543 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4544 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4549 len = ext4_blocks_count(es);
4550 start = sb_block + 1;
4551 brelse(bh); /* we're done with the superblock */
4553 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4554 start, len, blocksize);
4556 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4559 journal->j_private = sb;
4560 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4561 wait_on_buffer(journal->j_sb_buffer);
4562 if (!buffer_uptodate(journal->j_sb_buffer)) {
4563 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4566 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4567 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4568 "user (unsupported) - %d",
4569 be32_to_cpu(journal->j_superblock->s_nr_users));
4572 EXT4_SB(sb)->journal_bdev = bdev;
4573 ext4_init_journal_params(sb, journal);
4577 jbd2_journal_destroy(journal);
4579 ext4_blkdev_put(bdev);
4583 static int ext4_load_journal(struct super_block *sb,
4584 struct ext4_super_block *es,
4585 unsigned long journal_devnum)
4588 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4591 int really_read_only;
4593 BUG_ON(!ext4_has_feature_journal(sb));
4595 if (journal_devnum &&
4596 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4597 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4598 "numbers have changed");
4599 journal_dev = new_decode_dev(journal_devnum);
4601 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4603 really_read_only = bdev_read_only(sb->s_bdev);
4606 * Are we loading a blank journal or performing recovery after a
4607 * crash? For recovery, we need to check in advance whether we
4608 * can get read-write access to the device.
4610 if (ext4_has_feature_journal_needs_recovery(sb)) {
4611 if (sb->s_flags & MS_RDONLY) {
4612 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4613 "required on readonly filesystem");
4614 if (really_read_only) {
4615 ext4_msg(sb, KERN_ERR, "write access "
4616 "unavailable, cannot proceed");
4619 ext4_msg(sb, KERN_INFO, "write access will "
4620 "be enabled during recovery");
4624 if (journal_inum && journal_dev) {
4625 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4626 "and inode journals!");
4631 if (!(journal = ext4_get_journal(sb, journal_inum)))
4634 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4638 if (!(journal->j_flags & JBD2_BARRIER))
4639 ext4_msg(sb, KERN_INFO, "barriers disabled");
4641 if (!ext4_has_feature_journal_needs_recovery(sb))
4642 err = jbd2_journal_wipe(journal, !really_read_only);
4644 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4646 memcpy(save, ((char *) es) +
4647 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4648 err = jbd2_journal_load(journal);
4650 memcpy(((char *) es) + EXT4_S_ERR_START,
4651 save, EXT4_S_ERR_LEN);
4656 ext4_msg(sb, KERN_ERR, "error loading journal");
4657 jbd2_journal_destroy(journal);
4661 EXT4_SB(sb)->s_journal = journal;
4662 ext4_clear_journal_err(sb, es);
4664 if (!really_read_only && journal_devnum &&
4665 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4666 es->s_journal_dev = cpu_to_le32(journal_devnum);
4668 /* Make sure we flush the recovery flag to disk. */
4669 ext4_commit_super(sb, 1);
4675 static int ext4_commit_super(struct super_block *sb, int sync)
4677 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4678 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4683 if (block_device_ejected(sb))
4687 * If the file system is mounted read-only, don't update the
4688 * superblock write time. This avoids updating the superblock
4689 * write time when we are mounting the root file system
4690 * read/only but we need to replay the journal; at that point,
4691 * for people who are east of GMT and who make their clock
4692 * tick in localtime for Windows bug-for-bug compatibility,
4693 * the clock is set in the future, and this will cause e2fsck
4694 * to complain and force a full file system check.
4696 if (!(sb->s_flags & MS_RDONLY))
4697 es->s_wtime = cpu_to_le32(get_seconds());
4698 if (sb->s_bdev->bd_part)
4699 es->s_kbytes_written =
4700 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4701 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4702 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4704 es->s_kbytes_written =
4705 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4706 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4707 ext4_free_blocks_count_set(es,
4708 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4709 &EXT4_SB(sb)->s_freeclusters_counter)));
4710 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4711 es->s_free_inodes_count =
4712 cpu_to_le32(percpu_counter_sum_positive(
4713 &EXT4_SB(sb)->s_freeinodes_counter));
4714 BUFFER_TRACE(sbh, "marking dirty");
4715 ext4_superblock_csum_set(sb);
4718 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4720 * Oh, dear. A previous attempt to write the
4721 * superblock failed. This could happen because the
4722 * USB device was yanked out. Or it could happen to
4723 * be a transient write error and maybe the block will
4724 * be remapped. Nothing we can do but to retry the
4725 * write and hope for the best.
4727 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4728 "superblock detected");
4729 clear_buffer_write_io_error(sbh);
4730 set_buffer_uptodate(sbh);
4732 mark_buffer_dirty(sbh);
4735 error = __sync_dirty_buffer(sbh,
4736 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4740 error = buffer_write_io_error(sbh);
4742 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4744 clear_buffer_write_io_error(sbh);
4745 set_buffer_uptodate(sbh);
4752 * Have we just finished recovery? If so, and if we are mounting (or
4753 * remounting) the filesystem readonly, then we will end up with a
4754 * consistent fs on disk. Record that fact.
4756 static void ext4_mark_recovery_complete(struct super_block *sb,
4757 struct ext4_super_block *es)
4759 journal_t *journal = EXT4_SB(sb)->s_journal;
4761 if (!ext4_has_feature_journal(sb)) {
4762 BUG_ON(journal != NULL);
4765 jbd2_journal_lock_updates(journal);
4766 if (jbd2_journal_flush(journal) < 0)
4769 if (ext4_has_feature_journal_needs_recovery(sb) &&
4770 sb->s_flags & MS_RDONLY) {
4771 ext4_clear_feature_journal_needs_recovery(sb);
4772 ext4_commit_super(sb, 1);
4776 jbd2_journal_unlock_updates(journal);
4780 * If we are mounting (or read-write remounting) a filesystem whose journal
4781 * has recorded an error from a previous lifetime, move that error to the
4782 * main filesystem now.
4784 static void ext4_clear_journal_err(struct super_block *sb,
4785 struct ext4_super_block *es)
4791 BUG_ON(!ext4_has_feature_journal(sb));
4793 journal = EXT4_SB(sb)->s_journal;
4796 * Now check for any error status which may have been recorded in the
4797 * journal by a prior ext4_error() or ext4_abort()
4800 j_errno = jbd2_journal_errno(journal);
4804 errstr = ext4_decode_error(sb, j_errno, nbuf);
4805 ext4_warning(sb, "Filesystem error recorded "
4806 "from previous mount: %s", errstr);
4807 ext4_warning(sb, "Marking fs in need of filesystem check.");
4809 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4810 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4811 ext4_commit_super(sb, 1);
4813 jbd2_journal_clear_err(journal);
4814 jbd2_journal_update_sb_errno(journal);
4819 * Force the running and committing transactions to commit,
4820 * and wait on the commit.
4822 int ext4_force_commit(struct super_block *sb)
4826 if (sb->s_flags & MS_RDONLY)
4829 journal = EXT4_SB(sb)->s_journal;
4830 return ext4_journal_force_commit(journal);
4833 static int ext4_sync_fs(struct super_block *sb, int wait)
4837 bool needs_barrier = false;
4838 struct ext4_sb_info *sbi = EXT4_SB(sb);
4840 trace_ext4_sync_fs(sb, wait);
4841 flush_workqueue(sbi->rsv_conversion_wq);
4843 * Writeback quota in non-journalled quota case - journalled quota has
4846 dquot_writeback_dquots(sb, -1);
4848 * Data writeback is possible w/o journal transaction, so barrier must
4849 * being sent at the end of the function. But we can skip it if
4850 * transaction_commit will do it for us.
4852 if (sbi->s_journal) {
4853 target = jbd2_get_latest_transaction(sbi->s_journal);
4854 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4855 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4856 needs_barrier = true;
4858 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4860 ret = jbd2_log_wait_commit(sbi->s_journal,
4863 } else if (wait && test_opt(sb, BARRIER))
4864 needs_barrier = true;
4865 if (needs_barrier) {
4867 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4876 * LVM calls this function before a (read-only) snapshot is created. This
4877 * gives us a chance to flush the journal completely and mark the fs clean.
4879 * Note that only this function cannot bring a filesystem to be in a clean
4880 * state independently. It relies on upper layer to stop all data & metadata
4883 static int ext4_freeze(struct super_block *sb)
4888 if (sb->s_flags & MS_RDONLY)
4891 journal = EXT4_SB(sb)->s_journal;
4894 /* Now we set up the journal barrier. */
4895 jbd2_journal_lock_updates(journal);
4898 * Don't clear the needs_recovery flag if we failed to
4899 * flush the journal.
4901 error = jbd2_journal_flush(journal);
4905 /* Journal blocked and flushed, clear needs_recovery flag. */
4906 ext4_clear_feature_journal_needs_recovery(sb);
4909 error = ext4_commit_super(sb, 1);
4912 /* we rely on upper layer to stop further updates */
4913 jbd2_journal_unlock_updates(journal);
4918 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4919 * flag here, even though the filesystem is not technically dirty yet.
4921 static int ext4_unfreeze(struct super_block *sb)
4923 if (sb->s_flags & MS_RDONLY)
4926 if (EXT4_SB(sb)->s_journal) {
4927 /* Reset the needs_recovery flag before the fs is unlocked. */
4928 ext4_set_feature_journal_needs_recovery(sb);
4931 ext4_commit_super(sb, 1);
4936 * Structure to save mount options for ext4_remount's benefit
4938 struct ext4_mount_options {
4939 unsigned long s_mount_opt;
4940 unsigned long s_mount_opt2;
4943 unsigned long s_commit_interval;
4944 u32 s_min_batch_time, s_max_batch_time;
4947 char *s_qf_names[EXT4_MAXQUOTAS];
4951 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4953 struct ext4_super_block *es;
4954 struct ext4_sb_info *sbi = EXT4_SB(sb);
4955 unsigned long old_sb_flags;
4956 struct ext4_mount_options old_opts;
4957 int enable_quota = 0;
4959 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4964 char *orig_data = kstrdup(data, GFP_KERNEL);
4966 /* Store the original options */
4967 old_sb_flags = sb->s_flags;
4968 old_opts.s_mount_opt = sbi->s_mount_opt;
4969 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4970 old_opts.s_resuid = sbi->s_resuid;
4971 old_opts.s_resgid = sbi->s_resgid;
4972 old_opts.s_commit_interval = sbi->s_commit_interval;
4973 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4974 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4976 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4977 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4978 if (sbi->s_qf_names[i]) {
4979 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4981 if (!old_opts.s_qf_names[i]) {
4982 for (j = 0; j < i; j++)
4983 kfree(old_opts.s_qf_names[j]);
4988 old_opts.s_qf_names[i] = NULL;
4990 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4991 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4993 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4998 ext4_clamp_want_extra_isize(sb);
5000 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5001 test_opt(sb, JOURNAL_CHECKSUM)) {
5002 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5003 "during remount not supported; ignoring");
5004 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5007 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5008 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5009 ext4_msg(sb, KERN_ERR, "can't mount with "
5010 "both data=journal and delalloc");
5014 if (test_opt(sb, DIOREAD_NOLOCK)) {
5015 ext4_msg(sb, KERN_ERR, "can't mount with "
5016 "both data=journal and dioread_nolock");
5020 if (test_opt(sb, DAX)) {
5021 ext4_msg(sb, KERN_ERR, "can't mount with "
5022 "both data=journal and dax");
5028 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5029 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5030 "dax flag with busy inodes while remounting");
5031 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5034 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5035 ext4_abort(sb, "Abort forced by user");
5037 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5038 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5042 if (sbi->s_journal) {
5043 ext4_init_journal_params(sb, sbi->s_journal);
5044 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5047 if (*flags & MS_LAZYTIME)
5048 sb->s_flags |= MS_LAZYTIME;
5050 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
5051 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5056 if (*flags & MS_RDONLY) {
5057 err = sync_filesystem(sb);
5060 err = dquot_suspend(sb, -1);
5065 * First of all, the unconditional stuff we have to do
5066 * to disable replay of the journal when we next remount
5068 sb->s_flags |= MS_RDONLY;
5071 * OK, test if we are remounting a valid rw partition
5072 * readonly, and if so set the rdonly flag and then
5073 * mark the partition as valid again.
5075 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5076 (sbi->s_mount_state & EXT4_VALID_FS))
5077 es->s_state = cpu_to_le16(sbi->s_mount_state);
5080 ext4_mark_recovery_complete(sb, es);
5082 /* Make sure we can mount this feature set readwrite */
5083 if (ext4_has_feature_readonly(sb) ||
5084 !ext4_feature_set_ok(sb, 0)) {
5089 * Make sure the group descriptor checksums
5090 * are sane. If they aren't, refuse to remount r/w.
5092 for (g = 0; g < sbi->s_groups_count; g++) {
5093 struct ext4_group_desc *gdp =
5094 ext4_get_group_desc(sb, g, NULL);
5096 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5097 ext4_msg(sb, KERN_ERR,
5098 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5099 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5100 le16_to_cpu(gdp->bg_checksum));
5107 * If we have an unprocessed orphan list hanging
5108 * around from a previously readonly bdev mount,
5109 * require a full umount/remount for now.
5111 if (es->s_last_orphan) {
5112 ext4_msg(sb, KERN_WARNING, "Couldn't "
5113 "remount RDWR because of unprocessed "
5114 "orphan inode list. Please "
5115 "umount/remount instead");
5121 * Mounting a RDONLY partition read-write, so reread
5122 * and store the current valid flag. (It may have
5123 * been changed by e2fsck since we originally mounted
5127 ext4_clear_journal_err(sb, es);
5128 sbi->s_mount_state = le16_to_cpu(es->s_state);
5129 if (!ext4_setup_super(sb, es, 0))
5130 sb->s_flags &= ~MS_RDONLY;
5131 if (ext4_has_feature_mmp(sb))
5132 if (ext4_multi_mount_protect(sb,
5133 le64_to_cpu(es->s_mmp_block))) {
5142 * Reinitialize lazy itable initialization thread based on
5145 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5146 ext4_unregister_li_request(sb);
5148 ext4_group_t first_not_zeroed;
5149 first_not_zeroed = ext4_has_uninit_itable(sb);
5150 ext4_register_li_request(sb, first_not_zeroed);
5153 err = ext4_setup_system_zone(sb);
5157 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5158 ext4_commit_super(sb, 1);
5161 /* Release old quota file names */
5162 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5163 kfree(old_opts.s_qf_names[i]);
5165 if (sb_any_quota_suspended(sb))
5166 dquot_resume(sb, -1);
5167 else if (ext4_has_feature_quota(sb)) {
5168 err = ext4_enable_quotas(sb);
5175 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5176 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5181 sb->s_flags = old_sb_flags;
5182 sbi->s_mount_opt = old_opts.s_mount_opt;
5183 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5184 sbi->s_resuid = old_opts.s_resuid;
5185 sbi->s_resgid = old_opts.s_resgid;
5186 sbi->s_commit_interval = old_opts.s_commit_interval;
5187 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5188 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5190 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5191 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5192 kfree(sbi->s_qf_names[i]);
5193 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5201 static int ext4_statfs_project(struct super_block *sb,
5202 kprojid_t projid, struct kstatfs *buf)
5205 struct dquot *dquot;
5209 qid = make_kqid_projid(projid);
5210 dquot = dqget(sb, qid);
5212 return PTR_ERR(dquot);
5213 spin_lock(&dq_data_lock);
5215 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5216 dquot->dq_dqb.dqb_bsoftlimit :
5217 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5218 if (limit && buf->f_blocks > limit) {
5219 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5220 buf->f_blocks = limit;
5221 buf->f_bfree = buf->f_bavail =
5222 (buf->f_blocks > curblock) ?
5223 (buf->f_blocks - curblock) : 0;
5226 limit = dquot->dq_dqb.dqb_isoftlimit ?
5227 dquot->dq_dqb.dqb_isoftlimit :
5228 dquot->dq_dqb.dqb_ihardlimit;
5229 if (limit && buf->f_files > limit) {
5230 buf->f_files = limit;
5232 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5233 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5236 spin_unlock(&dq_data_lock);
5242 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5244 struct super_block *sb = dentry->d_sb;
5245 struct ext4_sb_info *sbi = EXT4_SB(sb);
5246 struct ext4_super_block *es = sbi->s_es;
5247 ext4_fsblk_t overhead = 0, resv_blocks;
5250 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5252 if (!test_opt(sb, MINIX_DF))
5253 overhead = sbi->s_overhead;
5255 buf->f_type = EXT4_SUPER_MAGIC;
5256 buf->f_bsize = sb->s_blocksize;
5257 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5258 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5259 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5260 /* prevent underflow in case that few free space is available */
5261 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5262 buf->f_bavail = buf->f_bfree -
5263 (ext4_r_blocks_count(es) + resv_blocks);
5264 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5266 buf->f_files = le32_to_cpu(es->s_inodes_count);
5267 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5268 buf->f_namelen = EXT4_NAME_LEN;
5269 fsid = le64_to_cpup((void *)es->s_uuid) ^
5270 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5271 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5272 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5275 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5276 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5277 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5282 /* Helper function for writing quotas on sync - we need to start transaction
5283 * before quota file is locked for write. Otherwise the are possible deadlocks:
5284 * Process 1 Process 2
5285 * ext4_create() quota_sync()
5286 * jbd2_journal_start() write_dquot()
5287 * dquot_initialize() down(dqio_mutex)
5288 * down(dqio_mutex) jbd2_journal_start()
5294 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5296 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5299 static int ext4_write_dquot(struct dquot *dquot)
5303 struct inode *inode;
5305 inode = dquot_to_inode(dquot);
5306 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5307 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5309 return PTR_ERR(handle);
5310 ret = dquot_commit(dquot);
5311 err = ext4_journal_stop(handle);
5317 static int ext4_acquire_dquot(struct dquot *dquot)
5322 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5323 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5325 return PTR_ERR(handle);
5326 ret = dquot_acquire(dquot);
5327 err = ext4_journal_stop(handle);
5333 static int ext4_release_dquot(struct dquot *dquot)
5338 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5339 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5340 if (IS_ERR(handle)) {
5341 /* Release dquot anyway to avoid endless cycle in dqput() */
5342 dquot_release(dquot);
5343 return PTR_ERR(handle);
5345 ret = dquot_release(dquot);
5346 err = ext4_journal_stop(handle);
5352 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5354 struct super_block *sb = dquot->dq_sb;
5355 struct ext4_sb_info *sbi = EXT4_SB(sb);
5357 /* Are we journaling quotas? */
5358 if (ext4_has_feature_quota(sb) ||
5359 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5360 dquot_mark_dquot_dirty(dquot);
5361 return ext4_write_dquot(dquot);
5363 return dquot_mark_dquot_dirty(dquot);
5367 static int ext4_write_info(struct super_block *sb, int type)
5372 /* Data block + inode block */
5373 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5375 return PTR_ERR(handle);
5376 ret = dquot_commit_info(sb, type);
5377 err = ext4_journal_stop(handle);
5384 * Turn on quotas during mount time - we need to find
5385 * the quota file and such...
5387 static int ext4_quota_on_mount(struct super_block *sb, int type)
5389 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5390 EXT4_SB(sb)->s_jquota_fmt, type);
5393 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5395 struct ext4_inode_info *ei = EXT4_I(inode);
5397 /* The first argument of lockdep_set_subclass has to be
5398 * *exactly* the same as the argument to init_rwsem() --- in
5399 * this case, in init_once() --- or lockdep gets unhappy
5400 * because the name of the lock is set using the
5401 * stringification of the argument to init_rwsem().
5403 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5404 lockdep_set_subclass(&ei->i_data_sem, subclass);
5408 * Standard function to be called on quota_on
5410 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5415 if (!test_opt(sb, QUOTA))
5418 /* Quotafile not on the same filesystem? */
5419 if (path->dentry->d_sb != sb)
5422 /* Quota already enabled for this file? */
5423 if (IS_NOQUOTA(d_inode(path->dentry)))
5426 /* Journaling quota? */
5427 if (EXT4_SB(sb)->s_qf_names[type]) {
5428 /* Quotafile not in fs root? */
5429 if (path->dentry->d_parent != sb->s_root)
5430 ext4_msg(sb, KERN_WARNING,
5431 "Quota file not on filesystem root. "
5432 "Journaled quota will not work");
5436 * When we journal data on quota file, we have to flush journal to see
5437 * all updates to the file when we bypass pagecache...
5439 if (EXT4_SB(sb)->s_journal &&
5440 ext4_should_journal_data(d_inode(path->dentry))) {
5442 * We don't need to lock updates but journal_flush() could
5443 * otherwise be livelocked...
5445 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5446 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5447 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5451 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5452 err = dquot_quota_on(sb, type, format_id, path);
5454 lockdep_set_quota_inode(path->dentry->d_inode,
5459 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5463 struct inode *qf_inode;
5464 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5465 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5466 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5467 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5470 BUG_ON(!ext4_has_feature_quota(sb));
5472 if (!qf_inums[type])
5475 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5476 if (IS_ERR(qf_inode)) {
5477 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5478 return PTR_ERR(qf_inode);
5481 /* Don't account quota for quota files to avoid recursion */
5482 qf_inode->i_flags |= S_NOQUOTA;
5483 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5484 err = dquot_enable(qf_inode, type, format_id, flags);
5486 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5492 /* Enable usage tracking for all quota types. */
5493 static int ext4_enable_quotas(struct super_block *sb)
5496 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5497 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5498 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5499 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5501 bool quota_mopt[EXT4_MAXQUOTAS] = {
5502 test_opt(sb, USRQUOTA),
5503 test_opt(sb, GRPQUOTA),
5504 test_opt(sb, PRJQUOTA),
5507 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5508 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5509 if (qf_inums[type]) {
5510 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5511 DQUOT_USAGE_ENABLED |
5512 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5514 for (type--; type >= 0; type--)
5515 dquot_quota_off(sb, type);
5518 "Failed to enable quota tracking "
5519 "(type=%d, err=%d). Please run "
5520 "e2fsck to fix.", type, err);
5528 static int ext4_quota_off(struct super_block *sb, int type)
5530 struct inode *inode = sb_dqopt(sb)->files[type];
5533 /* Force all delayed allocation blocks to be allocated.
5534 * Caller already holds s_umount sem */
5535 if (test_opt(sb, DELALLOC))
5536 sync_filesystem(sb);
5541 /* Update modification times of quota files when userspace can
5542 * start looking at them */
5543 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5546 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5547 ext4_mark_inode_dirty(handle, inode);
5548 ext4_journal_stop(handle);
5551 return dquot_quota_off(sb, type);
5554 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5555 * acquiring the locks... As quota files are never truncated and quota code
5556 * itself serializes the operations (and no one else should touch the files)
5557 * we don't have to be afraid of races */
5558 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5559 size_t len, loff_t off)
5561 struct inode *inode = sb_dqopt(sb)->files[type];
5562 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5563 int offset = off & (sb->s_blocksize - 1);
5566 struct buffer_head *bh;
5567 loff_t i_size = i_size_read(inode);
5571 if (off+len > i_size)
5574 while (toread > 0) {
5575 tocopy = sb->s_blocksize - offset < toread ?
5576 sb->s_blocksize - offset : toread;
5577 bh = ext4_bread(NULL, inode, blk, 0);
5580 if (!bh) /* A hole? */
5581 memset(data, 0, tocopy);
5583 memcpy(data, bh->b_data+offset, tocopy);
5593 /* Write to quotafile (we know the transaction is already started and has
5594 * enough credits) */
5595 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5596 const char *data, size_t len, loff_t off)
5598 struct inode *inode = sb_dqopt(sb)->files[type];
5599 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5600 int err, offset = off & (sb->s_blocksize - 1);
5602 struct buffer_head *bh;
5603 handle_t *handle = journal_current_handle();
5606 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5607 " cancelled because transaction is not started",
5608 (unsigned long long)off, (unsigned long long)len);
5612 * Since we account only one data block in transaction credits,
5613 * then it is impossible to cross a block boundary.
5615 if (sb->s_blocksize - offset < len) {
5616 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5617 " cancelled because not block aligned",
5618 (unsigned long long)off, (unsigned long long)len);
5623 bh = ext4_bread(handle, inode, blk,
5624 EXT4_GET_BLOCKS_CREATE |
5625 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5626 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5627 ext4_should_retry_alloc(inode->i_sb, &retries));
5632 BUFFER_TRACE(bh, "get write access");
5633 err = ext4_journal_get_write_access(handle, bh);
5639 memcpy(bh->b_data+offset, data, len);
5640 flush_dcache_page(bh->b_page);
5642 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5645 if (inode->i_size < off + len) {
5646 i_size_write(inode, off + len);
5647 EXT4_I(inode)->i_disksize = inode->i_size;
5648 ext4_mark_inode_dirty(handle, inode);
5653 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5655 const struct quota_format_ops *ops;
5657 if (!sb_has_quota_loaded(sb, qid->type))
5659 ops = sb_dqopt(sb)->ops[qid->type];
5660 if (!ops || !ops->get_next_id)
5662 return dquot_get_next_id(sb, qid);
5666 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5667 const char *dev_name, void *data)
5669 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5672 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5673 static inline void register_as_ext2(void)
5675 int err = register_filesystem(&ext2_fs_type);
5678 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5681 static inline void unregister_as_ext2(void)
5683 unregister_filesystem(&ext2_fs_type);
5686 static inline int ext2_feature_set_ok(struct super_block *sb)
5688 if (ext4_has_unknown_ext2_incompat_features(sb))
5690 if (sb->s_flags & MS_RDONLY)
5692 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5697 static inline void register_as_ext2(void) { }
5698 static inline void unregister_as_ext2(void) { }
5699 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5702 static inline void register_as_ext3(void)
5704 int err = register_filesystem(&ext3_fs_type);
5707 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5710 static inline void unregister_as_ext3(void)
5712 unregister_filesystem(&ext3_fs_type);
5715 static inline int ext3_feature_set_ok(struct super_block *sb)
5717 if (ext4_has_unknown_ext3_incompat_features(sb))
5719 if (!ext4_has_feature_journal(sb))
5721 if (sb->s_flags & MS_RDONLY)
5723 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5728 static struct file_system_type ext4_fs_type = {
5729 .owner = THIS_MODULE,
5731 .mount = ext4_mount,
5732 .kill_sb = kill_block_super,
5733 .fs_flags = FS_REQUIRES_DEV,
5735 MODULE_ALIAS_FS("ext4");
5737 /* Shared across all ext4 file systems */
5738 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5740 static int __init ext4_init_fs(void)
5744 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5745 ext4_li_info = NULL;
5746 mutex_init(&ext4_li_mtx);
5748 /* Build-time check for flags consistency */
5749 ext4_check_flag_values();
5751 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5752 init_waitqueue_head(&ext4__ioend_wq[i]);
5754 err = ext4_init_es();
5758 err = ext4_init_pageio();
5762 err = ext4_init_system_zone();
5766 err = ext4_init_sysfs();
5770 err = ext4_init_mballoc();
5773 err = init_inodecache();
5778 err = register_filesystem(&ext4_fs_type);
5784 unregister_as_ext2();
5785 unregister_as_ext3();
5786 destroy_inodecache();
5788 ext4_exit_mballoc();
5792 ext4_exit_system_zone();
5801 static void __exit ext4_exit_fs(void)
5803 ext4_destroy_lazyinit_thread();
5804 unregister_as_ext2();
5805 unregister_as_ext3();
5806 unregister_filesystem(&ext4_fs_type);
5807 destroy_inodecache();
5808 ext4_exit_mballoc();
5810 ext4_exit_system_zone();
5815 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5816 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5817 MODULE_LICENSE("GPL");
5818 module_init(ext4_init_fs)
5819 module_exit(ext4_exit_fs)